projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
don't put symlink bodies in pagecache into highmem
[linux-2.6-block.git] / fs / proc / base.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
e070ad49
ML		 14 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
1da177e4
LT		 48 */
49
50#include <asm/uaccess.h>
51
1da177e4
LT		 52#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
5995477a 56#include <linux/task_io_accounting_ops.h>
1da177e4 57#include <linux/init.h>
16f7e0fe 58#include <linux/capability.h>
1da177e4 59#include <linux/file.h>
9f3acc31 60#include <linux/fdtable.h>
1da177e4
LT		 61#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
6b3286ed 64#include <linux/mnt_namespace.h>
1da177e4 65#include <linux/mm.h>
a63d83f4 66#include <linux/swap.h>
b835996f 67#include <linux/rcupdate.h>
1da177e4 68#include <linux/kallsyms.h>
2ec220e2 69#include <linux/stacktrace.h>
d85f50d5 70#include <linux/resource.h>
5096add8 71#include <linux/module.h>
1da177e4
LT		 72#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
0d094efe 75#include <linux/tracehook.h>
87ebdc00 76#include <linux/printk.h>
a424316c 77#include <linux/cgroup.h>
1da177e4
LT		 78#include <linux/cpuset.h>
79#include <linux/audit.h>
5addc5dd 80#include <linux/poll.h>
1651e14e 81#include <linux/nsproxy.h>
8ac773b4 82#include <linux/oom.h>
3cb4a0bb 83#include <linux/elf.h>
60347f67 84#include <linux/pid_namespace.h>
22d917d8 85#include <linux/user_namespace.h>
5ad4e53b 86#include <linux/fs_struct.h>
5a0e3ad6 87#include <linux/slab.h>
640708a2 88#include <linux/flex_array.h>
48f6a7a5 89#include <linux/posix-timers.h>
f133ecca
CM		 90#ifdef CONFIG_HARDWALL
91#include <asm/hardwall.h>
92#endif
43d2b113 93#include <trace/events/oom.h>
1da177e4 94#include "internal.h"
faf60af1 95#include "fd.h"
1da177e4 96
0f2fe20f
EB		 97/* NOTE:
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
102 *
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
105 */
106
1da177e4 107struct pid_entry {
cedbccab 108 const char *name;
c5141e6d 109 int len;
d161a13f 110 umode_t mode;
c5ef1c42 111 const struct inode_operations *iop;
00977a59 112 const struct file_operations *fop;
20cdc894 113 union proc_op op;
1da177e4
LT		 114};
115
61a28784 116#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 117 .name = (NAME), \
c5141e6d 118 .len = sizeof(NAME) - 1, \
20cdc894
EB		 119 .mode = MODE, \
120 .iop = IOP, \
121 .fop = FOP, \
122 .op = OP, \
123}
124
631f9c18
AD		 125#define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127#define LNK(NAME, get_link) \
61a28784 128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 129 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 130 { .proc_get_link = get_link } )
131#define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
631f9c18 133#define ONE(NAME, MODE, show) \
be614086
EB		 134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
631f9c18 136 { .proc_show = show } )
1da177e4 137
aed54175
VN		 138/*
139 * Count the number of hardlinks for the pid_entry table, excluding the .
140 * and .. links.
141 */
142static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 unsigned int n)
144{
145 unsigned int i;
146 unsigned int count;
147
148 count = 0;
149 for (i = 0; i < n; ++i) {
150 if (S_ISDIR(entries[i].mode))
151 ++count;
152 }
153
154 return count;
155}
156
f7ad3c6b 157static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 158{
7c2c7d99
HD		 159 int result = -ENOENT;
160
0494f6ec 161 task_lock(task);
f7ad3c6b
MS		 162 if (task->fs) {
163 get_fs_root(task->fs, root);
7c2c7d99
HD		 164 result = 0;
165 }
0494f6ec 166 task_unlock(task);
7c2c7d99 167 return result;
0494f6ec
MS		 168}
169
7773fbc5 170static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 171{
2b0143b5 172 struct task_struct *task = get_proc_task(d_inode(dentry));
0494f6ec 173 int result = -ENOENT;
99f89551
EB		 174
175 if (task) {
f7ad3c6b
MS		 176 task_lock(task);
177 if (task->fs) {
178 get_fs_pwd(task->fs, path);
179 result = 0;
180 }
181 task_unlock(task);
99f89551
EB		 182 put_task_struct(task);
183 }
1da177e4
LT		 184 return result;
185}
186
7773fbc5 187static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 188{
2b0143b5 189 struct task_struct *task = get_proc_task(d_inode(dentry));
1da177e4 190 int result = -ENOENT;
99f89551
EB		 191
192 if (task) {
f7ad3c6b 193 result = get_task_root(task, path);
99f89551
EB		 194 put_task_struct(task);
195 }
1da177e4
LT		 196 return result;
197}
198
c2c0bb44
AD		 199static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
200 size_t _count, loff_t *pos)
1da177e4 201{
c2c0bb44
AD		 202 struct task_struct *tsk;
203 struct mm_struct *mm;
204 char *page;
205 unsigned long count = _count;
206 unsigned long arg_start, arg_end, env_start, env_end;
207 unsigned long len1, len2, len;
208 unsigned long p;
209 char c;
210 ssize_t rv;
211
212 BUG_ON(*pos < 0);
213
214 tsk = get_proc_task(file_inode(file));
215 if (!tsk)
216 return -ESRCH;
217 mm = get_task_mm(tsk);
218 put_task_struct(tsk);
219 if (!mm)
220 return 0;
221 /* Check if process spawned far enough to have cmdline. */
222 if (!mm->env_end) {
223 rv = 0;
224 goto out_mmput;
225 }
226
227 page = (char *)__get_free_page(GFP_TEMPORARY);
228 if (!page) {
229 rv = -ENOMEM;
230 goto out_mmput;
231 }
232
233 down_read(&mm->mmap_sem);
234 arg_start = mm->arg_start;
235 arg_end = mm->arg_end;
236 env_start = mm->env_start;
237 env_end = mm->env_end;
238 up_read(&mm->mmap_sem);
239
240 BUG_ON(arg_start > arg_end);
241 BUG_ON(env_start > env_end);
242
243 len1 = arg_end - arg_start;
244 len2 = env_end - env_start;
245
3581d458
AD		 246 /* Empty ARGV. */
247 if (len1 == 0) {
248 rv = 0;
249 goto out_free_page;
250 }
2ca66ff7 251 /*
c2c0bb44
AD		 252 * Inherently racy -- command line shares address space
253 * with code and data.
2ca66ff7 254 */
c2c0bb44
AD		 255 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
256 if (rv <= 0)
257 goto out_free_page;
258
259 rv = 0;
260
261 if (c == '\0') {
262 /* Command line (set of strings) occupies whole ARGV. */
263 if (len1 <= *pos)
264 goto out_free_page;
265
266 p = arg_start + *pos;
267 len = len1 - *pos;
268 while (count > 0 && len > 0) {
269 unsigned int _count;
270 int nr_read;
271
272 _count = min3(count, len, PAGE_SIZE);
273 nr_read = access_remote_vm(mm, p, page, _count, 0);
274 if (nr_read < 0)
275 rv = nr_read;
276 if (nr_read <= 0)
277 goto out_free_page;
278
279 if (copy_to_user(buf, page, nr_read)) {
280 rv = -EFAULT;
281 goto out_free_page;
282 }
283
284 p += nr_read;
285 len -= nr_read;
286 buf += nr_read;
287 count -= nr_read;
288 rv += nr_read;
289 }
290 } else {
291 /*
292 * Command line (1 string) occupies ARGV and maybe
293 * extends into ENVP.
294 */
295 if (len1 + len2 <= *pos)
296 goto skip_argv_envp;
297 if (len1 <= *pos)
298 goto skip_argv;
299
300 p = arg_start + *pos;
301 len = len1 - *pos;
302 while (count > 0 && len > 0) {
303 unsigned int _count, l;
304 int nr_read;
305 bool final;
306
307 _count = min3(count, len, PAGE_SIZE);
308 nr_read = access_remote_vm(mm, p, page, _count, 0);
309 if (nr_read < 0)
310 rv = nr_read;
311 if (nr_read <= 0)
312 goto out_free_page;
313
314 /*
315 * Command line can be shorter than whole ARGV
316 * even if last "marker" byte says it is not.
317 */
318 final = false;
319 l = strnlen(page, nr_read);
320 if (l < nr_read) {
321 nr_read = l;
322 final = true;
323 }
324
325 if (copy_to_user(buf, page, nr_read)) {
326 rv = -EFAULT;
327 goto out_free_page;
328 }
329
330 p += nr_read;
331 len -= nr_read;
332 buf += nr_read;
333 count -= nr_read;
334 rv += nr_read;
335
336 if (final)
337 goto out_free_page;
338 }
339skip_argv:
340 /*
341 * Command line (1 string) occupies ARGV and
342 * extends into ENVP.
343 */
344 if (len1 <= *pos) {
345 p = env_start + *pos - len1;
346 len = len1 + len2 - *pos;
347 } else {
348 p = env_start;
349 len = len2;
350 }
351 while (count > 0 && len > 0) {
352 unsigned int _count, l;
353 int nr_read;
354 bool final;
355
356 _count = min3(count, len, PAGE_SIZE);
357 nr_read = access_remote_vm(mm, p, page, _count, 0);
358 if (nr_read < 0)
359 rv = nr_read;
360 if (nr_read <= 0)
361 goto out_free_page;
362
363 /* Find EOS. */
364 final = false;
365 l = strnlen(page, nr_read);
366 if (l < nr_read) {
367 nr_read = l;
368 final = true;
369 }
370
371 if (copy_to_user(buf, page, nr_read)) {
372 rv = -EFAULT;
373 goto out_free_page;
374 }
375
376 p += nr_read;
377 len -= nr_read;
378 buf += nr_read;
379 count -= nr_read;
380 rv += nr_read;
381
382 if (final)
383 goto out_free_page;
384 }
385skip_argv_envp:
386 ;
387 }
388
389out_free_page:
390 free_page((unsigned long)page);
391out_mmput:
392 mmput(mm);
393 if (rv > 0)
394 *pos += rv;
395 return rv;
1da177e4
LT		 396}
397
c2c0bb44
AD		 398static const struct file_operations proc_pid_cmdline_ops = {
399 .read = proc_pid_cmdline_read,
400 .llseek = generic_file_llseek,
401};
402
f9ea536e
AD		 403static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
404 struct pid *pid, struct task_struct *task)
1da177e4 405{
e7dcd999 406 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
2fadaef4 407 if (mm && !IS_ERR(mm)) {
1da177e4 408 unsigned int nwords = 0;
dfe6b7d9 409 do {
1da177e4 410 nwords += 2;
dfe6b7d9 411 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
f9ea536e 412 seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
1da177e4 413 mmput(mm);
f9ea536e
AD		 414 return 0;
415 } else
416 return PTR_ERR(mm);
1da177e4
LT		 417}
418
419
420#ifdef CONFIG_KALLSYMS
421/*
422 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
423 * Returns the resolved symbol. If that fails, simply return the address.
424 */
edfcd606
AD		 425static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
426 struct pid *pid, struct task_struct *task)
1da177e4 427{
ffb45122 428 unsigned long wchan;
9281acea 429 char symname[KSYM_NAME_LEN];
1da177e4
LT		 430
431 wchan = get_wchan(task);
432
b2f73922 433 if (wchan && ptrace_may_access(task, PTRACE_MODE_READ) && !lookup_symbol_name(wchan, symname))
25ce3191 434 seq_printf(m, "%s", symname);
b2f73922
IM		 435 else
436 seq_putc(m, '0');
25ce3191
JP		 437
438 return 0;
1da177e4
LT		 439}
440#endif /* CONFIG_KALLSYMS */
441
a9712bc1
AV		 442static int lock_trace(struct task_struct *task)
443{
444 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
445 if (err)
446 return err;
447 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
448 mutex_unlock(&task->signal->cred_guard_mutex);
449 return -EPERM;
450 }
451 return 0;
452}
453
454static void unlock_trace(struct task_struct *task)
455{
456 mutex_unlock(&task->signal->cred_guard_mutex);
457}
458
2ec220e2
KC		 459#ifdef CONFIG_STACKTRACE
460
461#define MAX_STACK_TRACE_DEPTH 64
462
463static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
464 struct pid *pid, struct task_struct *task)
465{
466 struct stack_trace trace;
467 unsigned long *entries;
a9712bc1 468 int err;
2ec220e2
KC		 469 int i;
470
471 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
472 if (!entries)
473 return -ENOMEM;
474
475 trace.nr_entries = 0;
476 trace.max_entries = MAX_STACK_TRACE_DEPTH;
477 trace.entries = entries;
478 trace.skip = 0;
2ec220e2 479
a9712bc1
AV		 480 err = lock_trace(task);
481 if (!err) {
482 save_stack_trace_tsk(task, &trace);
483
484 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 485 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 486 (void *)entries[i], (void *)entries[i]);
487 }
488 unlock_trace(task);
2ec220e2
KC		 489 }
490 kfree(entries);
491
a9712bc1 492 return err;
2ec220e2
KC		 493}
494#endif
495
5968cece 496#ifdef CONFIG_SCHED_INFO
1da177e4
LT		 497/*
498 * Provides /proc/PID/schedstat
499 */
f6e826ca
AD		 500static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
501 struct pid *pid, struct task_struct *task)
1da177e4 502{
5968cece
NR		 503 if (unlikely(!sched_info_on()))
504 seq_printf(m, "0 0 0\n");
505 else
506 seq_printf(m, "%llu %llu %lu\n",
25ce3191
JP		 507 (unsigned long long)task->se.sum_exec_runtime,
508 (unsigned long long)task->sched_info.run_delay,
509 task->sched_info.pcount);
510
511 return 0;
1da177e4
LT		 512}
513#endif
514
9745512c
AV		 515#ifdef CONFIG_LATENCYTOP
516static int lstats_show_proc(struct seq_file *m, void *v)
517{
518 int i;
13d77c37
HS		 519 struct inode *inode = m->private;
520 struct task_struct *task = get_proc_task(inode);
9745512c 521
13d77c37
HS		 522 if (!task)
523 return -ESRCH;
524 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 525 for (i = 0; i < 32; i++) {
34e49d4f
JP		 526 struct latency_record *lr = &task->latency_record[i];
527 if (lr->backtrace[0]) {
9745512c 528 int q;
34e49d4f
JP		 529 seq_printf(m, "%i %li %li",
530 lr->count, lr->time, lr->max);
9745512c 531 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 532 unsigned long bt = lr->backtrace[q];
533 if (!bt)
9745512c 534 break;
34e49d4f 535 if (bt == ULONG_MAX)
9745512c 536 break;
34e49d4f 537 seq_printf(m, " %ps", (void *)bt);
9745512c 538 }
9d6de12f 539 seq_putc(m, '\n');
9745512c
AV		 540 }
541
542 }
13d77c37 543 put_task_struct(task);
9745512c
AV		 544 return 0;
545}
546
547static int lstats_open(struct inode *inode, struct file *file)
548{
13d77c37 549 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 550}
551
9745512c
AV		 552static ssize_t lstats_write(struct file *file, const char __user *buf,
553 size_t count, loff_t *offs)
554{
496ad9aa 555 struct task_struct *task = get_proc_task(file_inode(file));
9745512c 556
13d77c37
HS		 557 if (!task)
558 return -ESRCH;
9745512c 559 clear_all_latency_tracing(task);
13d77c37 560 put_task_struct(task);
9745512c
AV		 561
562 return count;
563}
564
565static const struct file_operations proc_lstats_operations = {
566 .open = lstats_open,
567 .read = seq_read,
568 .write = lstats_write,
569 .llseek = seq_lseek,
13d77c37 570 .release = single_release,
9745512c
AV		 571};
572
573#endif
574
6ba51e37
AD		 575static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
576 struct pid *pid, struct task_struct *task)
1da177e4 577{
a7f638f9 578 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 579 unsigned long points = 0;
1da177e4 580
19c5d45a 581 read_lock(&tasklist_lock);
b95c35e7 582 if (pid_alive(task))
a7f638f9
DR		 583 points = oom_badness(task, NULL, NULL, totalpages) *
584 1000 / totalpages;
19c5d45a 585 read_unlock(&tasklist_lock);
25ce3191
JP		 586 seq_printf(m, "%lu\n", points);
587
588 return 0;
1da177e4
LT		 589}
590
d85f50d5 591struct limit_names {
cedbccab
AD		 592 const char *name;
593 const char *unit;
d85f50d5
NH		 594};
595
596static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 597 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 598 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
599 [RLIMIT_DATA] = {"Max data size", "bytes"},
600 [RLIMIT_STACK] = {"Max stack size", "bytes"},
601 [RLIMIT_CORE] = {"Max core file size", "bytes"},
602 [RLIMIT_RSS] = {"Max resident set", "bytes"},
603 [RLIMIT_NPROC] = {"Max processes", "processes"},
604 [RLIMIT_NOFILE] = {"Max open files", "files"},
605 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
606 [RLIMIT_AS] = {"Max address space", "bytes"},
607 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
608 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
609 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
610 [RLIMIT_NICE] = {"Max nice priority", NULL},
611 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 612 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 613};
614
615/* Display limits for a process */
1c963eb1
AD		 616static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
617 struct pid *pid, struct task_struct *task)
d85f50d5
NH		 618{
619 unsigned int i;
d85f50d5 620 unsigned long flags;
d85f50d5
NH		 621
622 struct rlimit rlim[RLIM_NLIMITS];
623
a6bebbc8 624 if (!lock_task_sighand(task, &flags))
d85f50d5 625 return 0;
d85f50d5
NH		 626 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
627 unlock_task_sighand(task, &flags);
d85f50d5
NH		 628
629 /*
630 * print the file header
631 */
1c963eb1 632 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
25ce3191 633 "Limit", "Soft Limit", "Hard Limit", "Units");
d85f50d5
NH		 634
635 for (i = 0; i < RLIM_NLIMITS; i++) {
636 if (rlim[i].rlim_cur == RLIM_INFINITY)
1c963eb1 637 seq_printf(m, "%-25s %-20s ",
25ce3191 638 lnames[i].name, "unlimited");
d85f50d5 639 else
1c963eb1 640 seq_printf(m, "%-25s %-20lu ",
25ce3191 641 lnames[i].name, rlim[i].rlim_cur);
d85f50d5
NH		 642
643 if (rlim[i].rlim_max == RLIM_INFINITY)
1c963eb1 644 seq_printf(m, "%-20s ", "unlimited");
d85f50d5 645 else
1c963eb1 646 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
d85f50d5
NH		 647
648 if (lnames[i].unit)
1c963eb1 649 seq_printf(m, "%-10s\n", lnames[i].unit);
d85f50d5 650 else
1c963eb1 651 seq_putc(m, '\n');
d85f50d5
NH		 652 }
653
1c963eb1 654 return 0;
d85f50d5
NH		 655}
656
ebcb6734 657#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6
AD		 658static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
659 struct pid *pid, struct task_struct *task)
ebcb6734
RM		 660{
661 long nr;
662 unsigned long args[6], sp, pc;
25ce3191
JP		 663 int res;
664
665 res = lock_trace(task);
a9712bc1
AV		 666 if (res)
667 return res;
ebcb6734
RM		 668
669 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
09d93bd6 670 seq_puts(m, "running\n");
a9712bc1 671 else if (nr < 0)
09d93bd6 672 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
a9712bc1 673 else
09d93bd6 674 seq_printf(m,
ebcb6734
RM		 675 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
676 nr,
677 args[0], args[1], args[2], args[3], args[4], args[5],
678 sp, pc);
a9712bc1 679 unlock_trace(task);
25ce3191
JP		 680
681 return 0;
ebcb6734
RM		 682}
683#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
684
1da177e4
LT		 685/************************************************************************/
686/* Here the fs part begins */
687/************************************************************************/
688
689/* permission checks */
778c1144 690static int proc_fd_access_allowed(struct inode *inode)
1da177e4 691{
778c1144
EB		 692 struct task_struct *task;
693 int allowed = 0;
df26c40e
EB		 694 /* Allow access to a task's file descriptors if it is us or we
695 * may use ptrace attach to the process and find out that
696 * information.
778c1144
EB		 697 */
698 task = get_proc_task(inode);
df26c40e 699 if (task) {
006ebb40 700 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 701 put_task_struct(task);
df26c40e 702 }
778c1144 703 return allowed;
1da177e4
LT		 704}
705
6b4e306a 706int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 707{
708 int error;
2b0143b5 709 struct inode *inode = d_inode(dentry);
6d76fa58
LT		 710
711 if (attr->ia_valid & ATTR_MODE)
712 return -EPERM;
713
714 error = inode_change_ok(inode, attr);
1025774c
CH		 715 if (error)
716 return error;
717
1025774c
CH		 718 setattr_copy(inode, attr);
719 mark_inode_dirty(inode);
720 return 0;
6d76fa58
LT		 721}
722
0499680a
VK		 723/*
724 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
725 * or euid/egid (for hide_pid_min=2)?
726 */
727static bool has_pid_permissions(struct pid_namespace *pid,
728 struct task_struct *task,
729 int hide_pid_min)
730{
731 if (pid->hide_pid < hide_pid_min)
732 return true;
733 if (in_group_p(pid->pid_gid))
734 return true;
735 return ptrace_may_access(task, PTRACE_MODE_READ);
736}
737
738
739static int proc_pid_permission(struct inode *inode, int mask)
740{
741 struct pid_namespace *pid = inode->i_sb->s_fs_info;
742 struct task_struct *task;
743 bool has_perms;
744
745 task = get_proc_task(inode);
a2ef990a
XF		 746 if (!task)
747 return -ESRCH;
0499680a
VK		 748 has_perms = has_pid_permissions(pid, task, 1);
749 put_task_struct(task);
750
751 if (!has_perms) {
752 if (pid->hide_pid == 2) {
753 /*
754 * Let's make getdents(), stat(), and open()
755 * consistent with each other. If a process
756 * may not stat() a file, it shouldn't be seen
757 * in procfs at all.
758 */
759 return -ENOENT;
760 }
761
762 return -EPERM;
763 }
764 return generic_permission(inode, mask);
765}
766
767
768
c5ef1c42 769static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 770 .setattr = proc_setattr,
771};
772
be614086
EB		 773static int proc_single_show(struct seq_file *m, void *v)
774{
775 struct inode *inode = m->private;
776 struct pid_namespace *ns;
777 struct pid *pid;
778 struct task_struct *task;
779 int ret;
780
781 ns = inode->i_sb->s_fs_info;
782 pid = proc_pid(inode);
783 task = get_pid_task(pid, PIDTYPE_PID);
784 if (!task)
785 return -ESRCH;
786
787 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
788
789 put_task_struct(task);
790 return ret;
791}
792
793static int proc_single_open(struct inode *inode, struct file *filp)
794{
c6a34058 795 return single_open(filp, proc_single_show, inode);
be614086
EB		 796}
797
798static const struct file_operations proc_single_file_operations = {
799 .open = proc_single_open,
800 .read = seq_read,
801 .llseek = seq_lseek,
802 .release = single_release,
803};
804
5381e169
ON		 805
806struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
1da177e4 807{
5381e169
ON		 808 struct task_struct *task = get_proc_task(inode);
809 struct mm_struct *mm = ERR_PTR(-ESRCH);
e268337d 810
5381e169
ON		 811 if (task) {
812 mm = mm_access(task, mode);
813 put_task_struct(task);
e268337d 814
5381e169
ON		 815 if (!IS_ERR_OR_NULL(mm)) {
816 /* ensure this mm_struct can't be freed */
817 atomic_inc(&mm->mm_count);
818 /* but do not pin its memory */
819 mmput(mm);
820 }
821 }
822
823 return mm;
824}
825
826static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
827{
828 struct mm_struct *mm = proc_mem_open(inode, mode);
e268337d
LT		 829
830 if (IS_ERR(mm))
831 return PTR_ERR(mm);
832
e268337d 833 file->private_data = mm;
1da177e4
LT		 834 return 0;
835}
836
b409e578
CW		 837static int mem_open(struct inode *inode, struct file *file)
838{
bc452b4b
DH		 839 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
840
841 /* OK to pass negative loff_t, we can catch out-of-range */
842 file->f_mode |= FMODE_UNSIGNED_OFFSET;
843
844 return ret;
b409e578
CW		 845}
846
572d34b9
ON		 847static ssize_t mem_rw(struct file *file, char __user *buf,
848 size_t count, loff_t *ppos, int write)
1da177e4 849{
e268337d 850 struct mm_struct *mm = file->private_data;
572d34b9
ON		 851 unsigned long addr = *ppos;
852 ssize_t copied;
1da177e4 853 char *page;
1da177e4 854
e268337d
LT		 855 if (!mm)
856 return 0;
99f89551 857
30cd8903
KM		 858 page = (char *)__get_free_page(GFP_TEMPORARY);
859 if (!page)
e268337d 860 return -ENOMEM;
1da177e4 861
f7ca54f4 862 copied = 0;
6d08f2c7
ON		 863 if (!atomic_inc_not_zero(&mm->mm_users))
864 goto free;
865
1da177e4 866 while (count > 0) {
572d34b9 867 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 868
572d34b9 869 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 870 copied = -EFAULT;
871 break;
872 }
572d34b9
ON		 873
874 this_len = access_remote_vm(mm, addr, page, this_len, write);
875 if (!this_len) {
1da177e4
LT		 876 if (!copied)
877 copied = -EIO;
878 break;
879 }
572d34b9
ON		 880
881 if (!write && copy_to_user(buf, page, this_len)) {
882 copied = -EFAULT;
883 break;
884 }
885
886 buf += this_len;
887 addr += this_len;
888 copied += this_len;
889 count -= this_len;
1da177e4 890 }
572d34b9 891 *ppos = addr;
30cd8903 892
6d08f2c7
ON		 893 mmput(mm);
894free:
30cd8903 895 free_page((unsigned long) page);
1da177e4
LT		 896 return copied;
897}
1da177e4 898
572d34b9
ON		 899static ssize_t mem_read(struct file *file, char __user *buf,
900 size_t count, loff_t *ppos)
901{
902 return mem_rw(file, buf, count, ppos, 0);
903}
904
905static ssize_t mem_write(struct file *file, const char __user *buf,
906 size_t count, loff_t *ppos)
907{
908 return mem_rw(file, (char __user*)buf, count, ppos, 1);
909}
910
85863e47 911loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 912{
913 switch (orig) {
914 case 0:
915 file->f_pos = offset;
916 break;
917 case 1:
918 file->f_pos += offset;
919 break;
920 default:
921 return -EINVAL;
922 }
923 force_successful_syscall_return();
924 return file->f_pos;
925}
926
e268337d
LT		 927static int mem_release(struct inode *inode, struct file *file)
928{
929 struct mm_struct *mm = file->private_data;
71879d3c 930 if (mm)
6d08f2c7 931 mmdrop(mm);
e268337d
LT		 932 return 0;
933}
934
00977a59 935static const struct file_operations proc_mem_operations = {
1da177e4
LT		 936 .llseek = mem_lseek,
937 .read = mem_read,
938 .write = mem_write,
939 .open = mem_open,
e268337d 940 .release = mem_release,
1da177e4
LT		 941};
942
b409e578
CW		 943static int environ_open(struct inode *inode, struct file *file)
944{
945 return __mem_open(inode, file, PTRACE_MODE_READ);
946}
947
315e28c8
JP		 948static ssize_t environ_read(struct file *file, char __user *buf,
949 size_t count, loff_t *ppos)
950{
315e28c8
JP		 951 char *page;
952 unsigned long src = *ppos;
b409e578
CW		 953 int ret = 0;
954 struct mm_struct *mm = file->private_data;
315e28c8 955
b409e578
CW		 956 if (!mm)
957 return 0;
315e28c8 958
315e28c8
JP		 959 page = (char *)__get_free_page(GFP_TEMPORARY);
960 if (!page)
b409e578 961 return -ENOMEM;
315e28c8 962
d6f64b89 963 ret = 0;
b409e578
CW		 964 if (!atomic_inc_not_zero(&mm->mm_users))
965 goto free;
315e28c8 966 while (count > 0) {
e8905ec2
DH		 967 size_t this_len, max_len;
968 int retval;
315e28c8 969
e8905ec2 970 if (src >= (mm->env_end - mm->env_start))
315e28c8
JP		 971 break;
972
e8905ec2
DH		 973 this_len = mm->env_end - (mm->env_start + src);
974
975 max_len = min_t(size_t, PAGE_SIZE, count);
976 this_len = min(max_len, this_len);
315e28c8 977
b409e578 978 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP		 979 page, this_len, 0);
980
981 if (retval <= 0) {
982 ret = retval;
983 break;
984 }
985
986 if (copy_to_user(buf, page, retval)) {
987 ret = -EFAULT;
988 break;
989 }
990
991 ret += retval;
992 src += retval;
993 buf += retval;
994 count -= retval;
995 }
996 *ppos = src;
315e28c8 997 mmput(mm);
b409e578
CW		 998
999free:
315e28c8 1000 free_page((unsigned long) page);
315e28c8
JP		 1001 return ret;
1002}
1003
1004static const struct file_operations proc_environ_operations = {
b409e578 1005 .open = environ_open,
315e28c8 1006 .read = environ_read,
87df8424 1007 .llseek = generic_file_llseek,
b409e578 1008 .release = mem_release,
315e28c8
JP		 1009};
1010
fa0cbbf1
DR		 1011static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1012 loff_t *ppos)
1013{
496ad9aa 1014 struct task_struct *task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1015 char buffer[PROC_NUMBUF];
1016 int oom_adj = OOM_ADJUST_MIN;
1017 size_t len;
1018 unsigned long flags;
1019
1020 if (!task)
1021 return -ESRCH;
1022 if (lock_task_sighand(task, &flags)) {
1023 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1024 oom_adj = OOM_ADJUST_MAX;
1025 else
1026 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1027 OOM_SCORE_ADJ_MAX;
1028 unlock_task_sighand(task, &flags);
1029 }
1030 put_task_struct(task);
1031 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1032 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1033}
1034
b72bdfa7
DR		 1035/*
1036 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1037 * kernels. The effective policy is defined by oom_score_adj, which has a
1038 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1039 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1040 * Processes that become oom disabled via oom_adj will still be oom disabled
1041 * with this implementation.
1042 *
1043 * oom_adj cannot be removed since existing userspace binaries use it.
1044 */
fa0cbbf1
DR		 1045static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1046 size_t count, loff_t *ppos)
1047{
1048 struct task_struct *task;
1049 char buffer[PROC_NUMBUF];
1050 int oom_adj;
1051 unsigned long flags;
1052 int err;
1053
1054 memset(buffer, 0, sizeof(buffer));
1055 if (count > sizeof(buffer) - 1)
1056 count = sizeof(buffer) - 1;
1057 if (copy_from_user(buffer, buf, count)) {
1058 err = -EFAULT;
1059 goto out;
1060 }
1061
1062 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1063 if (err)
1064 goto out;
1065 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1066 oom_adj != OOM_DISABLE) {
1067 err = -EINVAL;
1068 goto out;
1069 }
1070
496ad9aa 1071 task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1072 if (!task) {
1073 err = -ESRCH;
1074 goto out;
1075 }
1076
1077 task_lock(task);
1078 if (!task->mm) {
1079 err = -EINVAL;
1080 goto err_task_lock;
1081 }
1082
1083 if (!lock_task_sighand(task, &flags)) {
1084 err = -ESRCH;
1085 goto err_task_lock;
1086 }
1087
1088 /*
1089 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1090 * value is always attainable.
1091 */
1092 if (oom_adj == OOM_ADJUST_MAX)
1093 oom_adj = OOM_SCORE_ADJ_MAX;
1094 else
1095 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1096
1097 if (oom_adj < task->signal->oom_score_adj &&
1098 !capable(CAP_SYS_RESOURCE)) {
1099 err = -EACCES;
1100 goto err_sighand;
1101 }
1102
1103 /*
1104 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1105 * /proc/pid/oom_score_adj instead.
1106 */
87ebdc00 1107 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
fa0cbbf1
DR		 1108 current->comm, task_pid_nr(current), task_pid_nr(task),
1109 task_pid_nr(task));
1110
1111 task->signal->oom_score_adj = oom_adj;
1112 trace_oom_score_adj_update(task);
1113err_sighand:
1114 unlock_task_sighand(task, &flags);
1115err_task_lock:
1116 task_unlock(task);
1117 put_task_struct(task);
1118out:
1119 return err < 0 ? err : count;
1120}
1121
1122static const struct file_operations proc_oom_adj_operations = {
1123 .read = oom_adj_read,
1124 .write = oom_adj_write,
1125 .llseek = generic_file_llseek,
1126};
1127
a63d83f4
DR		 1128static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1129 size_t count, loff_t *ppos)
1130{
496ad9aa 1131 struct task_struct *task = get_proc_task(file_inode(file));
a63d83f4 1132 char buffer[PROC_NUMBUF];
a9c58b90 1133 short oom_score_adj = OOM_SCORE_ADJ_MIN;
a63d83f4
DR		 1134 unsigned long flags;
1135 size_t len;
1136
1137 if (!task)
1138 return -ESRCH;
1139 if (lock_task_sighand(task, &flags)) {
1140 oom_score_adj = task->signal->oom_score_adj;
1141 unlock_task_sighand(task, &flags);
1142 }
1143 put_task_struct(task);
a9c58b90 1144 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
a63d83f4
DR		 1145 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1146}
1147
1148static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1149 size_t count, loff_t *ppos)
1150{
1151 struct task_struct *task;
1152 char buffer[PROC_NUMBUF];
1153 unsigned long flags;
0a8cb8e3 1154 int oom_score_adj;
a63d83f4
DR		 1155 int err;
1156
1157 memset(buffer, 0, sizeof(buffer));
1158 if (count > sizeof(buffer) - 1)
1159 count = sizeof(buffer) - 1;
723548bf
DR		 1160 if (copy_from_user(buffer, buf, count)) {
1161 err = -EFAULT;
1162 goto out;
1163 }
a63d83f4 1164
0a8cb8e3 1165 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1166 if (err)
723548bf 1167 goto out;
a63d83f4 1168 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1169 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1170 err = -EINVAL;
1171 goto out;
1172 }
a63d83f4 1173
496ad9aa 1174 task = get_proc_task(file_inode(file));
723548bf
DR		 1175 if (!task) {
1176 err = -ESRCH;
1177 goto out;
1178 }
d19d5476
DR		 1179
1180 task_lock(task);
1181 if (!task->mm) {
1182 err = -EINVAL;
1183 goto err_task_lock;
1184 }
1185
a63d83f4 1186 if (!lock_task_sighand(task, &flags)) {
723548bf 1187 err = -ESRCH;
d19d5476 1188 goto err_task_lock;
a63d83f4 1189 }
d19d5476 1190
a9c58b90 1191 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1192 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1193 err = -EACCES;
1194 goto err_sighand;
a63d83f4
DR		 1195 }
1196
a9c58b90 1197 task->signal->oom_score_adj = (short)oom_score_adj;
dabb16f6 1198 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
a9c58b90 1199 task->signal->oom_score_adj_min = (short)oom_score_adj;
43d2b113 1200 trace_oom_score_adj_update(task);
01dc52eb 1201
723548bf 1202err_sighand:
a63d83f4 1203 unlock_task_sighand(task, &flags);
d19d5476
DR		 1204err_task_lock:
1205 task_unlock(task);
a63d83f4 1206 put_task_struct(task);
723548bf
DR		 1207out:
1208 return err < 0 ? err : count;
a63d83f4
DR		 1209}
1210
1211static const struct file_operations proc_oom_score_adj_operations = {
1212 .read = oom_score_adj_read,
1213 .write = oom_score_adj_write,
6038f373 1214 .llseek = default_llseek,
a63d83f4
DR		 1215};
1216
1da177e4
LT		 1217#ifdef CONFIG_AUDITSYSCALL
1218#define TMPBUFLEN 21
1219static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1220 size_t count, loff_t *ppos)
1221{
496ad9aa 1222 struct inode * inode = file_inode(file);
99f89551 1223 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1224 ssize_t length;
1225 char tmpbuf[TMPBUFLEN];
1226
99f89551
EB		 1227 if (!task)
1228 return -ESRCH;
1da177e4 1229 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
e1760bd5
EB		 1230 from_kuid(file->f_cred->user_ns,
1231 audit_get_loginuid(task)));
99f89551 1232 put_task_struct(task);
1da177e4
LT		 1233 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1234}
1235
1236static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1237 size_t count, loff_t *ppos)
1238{
496ad9aa 1239 struct inode * inode = file_inode(file);
1da177e4 1240 uid_t loginuid;
e1760bd5 1241 kuid_t kloginuid;
774636e1 1242 int rv;
1da177e4 1243
7dc52157
PM		 1244 rcu_read_lock();
1245 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1246 rcu_read_unlock();
1da177e4 1247 return -EPERM;
7dc52157
PM		 1248 }
1249 rcu_read_unlock();
1da177e4 1250
1da177e4
LT		 1251 if (*ppos != 0) {
1252 /* No partial writes. */
1253 return -EINVAL;
1254 }
1da177e4 1255
774636e1
AD		 1256 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1257 if (rv < 0)
1258 return rv;
81407c84
EP		 1259
1260 /* is userspace tring to explicitly UNSET the loginuid? */
1261 if (loginuid == AUDIT_UID_UNSET) {
1262 kloginuid = INVALID_UID;
1263 } else {
1264 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
774636e1
AD		 1265 if (!uid_valid(kloginuid))
1266 return -EINVAL;
e1760bd5
EB		 1267 }
1268
774636e1
AD		 1269 rv = audit_set_loginuid(kloginuid);
1270 if (rv < 0)
1271 return rv;
1272 return count;
1da177e4
LT		 1273}
1274
00977a59 1275static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1276 .read = proc_loginuid_read,
1277 .write = proc_loginuid_write,
87df8424 1278 .llseek = generic_file_llseek,
1da177e4 1279};
1e0bd755
EP		 1280
1281static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1282 size_t count, loff_t *ppos)
1283{
496ad9aa 1284 struct inode * inode = file_inode(file);
1e0bd755
EP		 1285 struct task_struct *task = get_proc_task(inode);
1286 ssize_t length;
1287 char tmpbuf[TMPBUFLEN];
1288
1289 if (!task)
1290 return -ESRCH;
1291 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1292 audit_get_sessionid(task));
1293 put_task_struct(task);
1294 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1295}
1296
1297static const struct file_operations proc_sessionid_operations = {
1298 .read = proc_sessionid_read,
87df8424 1299 .llseek = generic_file_llseek,
1e0bd755 1300};
1da177e4
LT		 1301#endif
1302
f4f154fd
AM		 1303#ifdef CONFIG_FAULT_INJECTION
1304static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1305 size_t count, loff_t *ppos)
1306{
496ad9aa 1307 struct task_struct *task = get_proc_task(file_inode(file));
f4f154fd
AM		 1308 char buffer[PROC_NUMBUF];
1309 size_t len;
1310 int make_it_fail;
f4f154fd
AM		 1311
1312 if (!task)
1313 return -ESRCH;
1314 make_it_fail = task->make_it_fail;
1315 put_task_struct(task);
1316
1317 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1318
1319 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1320}
1321
1322static ssize_t proc_fault_inject_write(struct file * file,
1323 const char __user * buf, size_t count, loff_t *ppos)
1324{
1325 struct task_struct *task;
774636e1 1326 char buffer[PROC_NUMBUF];
f4f154fd 1327 int make_it_fail;
774636e1 1328 int rv;
f4f154fd
AM		 1329
1330 if (!capable(CAP_SYS_RESOURCE))
1331 return -EPERM;
1332 memset(buffer, 0, sizeof(buffer));
1333 if (count > sizeof(buffer) - 1)
1334 count = sizeof(buffer) - 1;
1335 if (copy_from_user(buffer, buf, count))
1336 return -EFAULT;
774636e1
AD		 1337 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1338 if (rv < 0)
1339 return rv;
16caed31
DJ		 1340 if (make_it_fail < 0 || make_it_fail > 1)
1341 return -EINVAL;
1342
496ad9aa 1343 task = get_proc_task(file_inode(file));
f4f154fd
AM		 1344 if (!task)
1345 return -ESRCH;
1346 task->make_it_fail = make_it_fail;
1347 put_task_struct(task);
cba8aafe
VL		 1348
1349 return count;
f4f154fd
AM		 1350}
1351
00977a59 1352static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1353 .read = proc_fault_inject_read,
1354 .write = proc_fault_inject_write,
87df8424 1355 .llseek = generic_file_llseek,
f4f154fd
AM		 1356};
1357#endif
1358
9745512c 1359
43ae34cb
IM		 1360#ifdef CONFIG_SCHED_DEBUG
1361/*
1362 * Print out various scheduling related per-task fields:
1363 */
1364static int sched_show(struct seq_file *m, void *v)
1365{
1366 struct inode *inode = m->private;
1367 struct task_struct *p;
1368
43ae34cb
IM		 1369 p = get_proc_task(inode);
1370 if (!p)
1371 return -ESRCH;
1372 proc_sched_show_task(p, m);
1373
1374 put_task_struct(p);
1375
1376 return 0;
1377}
1378
1379static ssize_t
1380sched_write(struct file *file, const char __user *buf,
1381 size_t count, loff_t *offset)
1382{
496ad9aa 1383 struct inode *inode = file_inode(file);
43ae34cb
IM		 1384 struct task_struct *p;
1385
43ae34cb
IM		 1386 p = get_proc_task(inode);
1387 if (!p)
1388 return -ESRCH;
1389 proc_sched_set_task(p);
1390
1391 put_task_struct(p);
1392
1393 return count;
1394}
1395
1396static int sched_open(struct inode *inode, struct file *filp)
1397{
c6a34058 1398 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1399}
1400
1401static const struct file_operations proc_pid_sched_operations = {
1402 .open = sched_open,
1403 .read = seq_read,
1404 .write = sched_write,
1405 .llseek = seq_lseek,
5ea473a1 1406 .release = single_release,
43ae34cb
IM		 1407};
1408
1409#endif
1410
5091faa4
MG		 1411#ifdef CONFIG_SCHED_AUTOGROUP
1412/*
1413 * Print out autogroup related information:
1414 */
1415static int sched_autogroup_show(struct seq_file *m, void *v)
1416{
1417 struct inode *inode = m->private;
1418 struct task_struct *p;
1419
1420 p = get_proc_task(inode);
1421 if (!p)
1422 return -ESRCH;
1423 proc_sched_autogroup_show_task(p, m);
1424
1425 put_task_struct(p);
1426
1427 return 0;
1428}
1429
1430static ssize_t
1431sched_autogroup_write(struct file *file, const char __user *buf,
1432 size_t count, loff_t *offset)
1433{
496ad9aa 1434 struct inode *inode = file_inode(file);
5091faa4
MG		 1435 struct task_struct *p;
1436 char buffer[PROC_NUMBUF];
0a8cb8e3 1437 int nice;
5091faa4
MG		 1438 int err;
1439
1440 memset(buffer, 0, sizeof(buffer));
1441 if (count > sizeof(buffer) - 1)
1442 count = sizeof(buffer) - 1;
1443 if (copy_from_user(buffer, buf, count))
1444 return -EFAULT;
1445
0a8cb8e3
AD		 1446 err = kstrtoint(strstrip(buffer), 0, &nice);
1447 if (err < 0)
1448 return err;
5091faa4
MG		 1449
1450 p = get_proc_task(inode);
1451 if (!p)
1452 return -ESRCH;
1453
2e5b5b3a 1454 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1455 if (err)
1456 count = err;
1457
1458 put_task_struct(p);
1459
1460 return count;
1461}
1462
1463static int sched_autogroup_open(struct inode *inode, struct file *filp)
1464{
1465 int ret;
1466
1467 ret = single_open(filp, sched_autogroup_show, NULL);
1468 if (!ret) {
1469 struct seq_file *m = filp->private_data;
1470
1471 m->private = inode;
1472 }
1473 return ret;
1474}
1475
1476static const struct file_operations proc_pid_sched_autogroup_operations = {
1477 .open = sched_autogroup_open,
1478 .read = seq_read,
1479 .write = sched_autogroup_write,
1480 .llseek = seq_lseek,
1481 .release = single_release,
1482};
1483
1484#endif /* CONFIG_SCHED_AUTOGROUP */
1485
4614a696 1486static ssize_t comm_write(struct file *file, const char __user *buf,
1487 size_t count, loff_t *offset)
1488{
496ad9aa 1489 struct inode *inode = file_inode(file);
4614a696 1490 struct task_struct *p;
1491 char buffer[TASK_COMM_LEN];
830e0fc9 1492 const size_t maxlen = sizeof(buffer) - 1;
4614a696 1493
1494 memset(buffer, 0, sizeof(buffer));
830e0fc9 1495 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
4614a696 1496 return -EFAULT;
1497
1498 p = get_proc_task(inode);
1499 if (!p)
1500 return -ESRCH;
1501
1502 if (same_thread_group(current, p))
1503 set_task_comm(p, buffer);
1504 else
1505 count = -EINVAL;
1506
1507 put_task_struct(p);
1508
1509 return count;
1510}
1511
1512static int comm_show(struct seq_file *m, void *v)
1513{
1514 struct inode *inode = m->private;
1515 struct task_struct *p;
1516
1517 p = get_proc_task(inode);
1518 if (!p)
1519 return -ESRCH;
1520
1521 task_lock(p);
1522 seq_printf(m, "%s\n", p->comm);
1523 task_unlock(p);
1524
1525 put_task_struct(p);
1526
1527 return 0;
1528}
1529
1530static int comm_open(struct inode *inode, struct file *filp)
1531{
c6a34058 1532 return single_open(filp, comm_show, inode);
4614a696 1533}
1534
1535static const struct file_operations proc_pid_set_comm_operations = {
1536 .open = comm_open,
1537 .read = seq_read,
1538 .write = comm_write,
1539 .llseek = seq_lseek,
1540 .release = single_release,
1541};
1542
7773fbc5 1543static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1544{
1545 struct task_struct *task;
1546 struct mm_struct *mm;
1547 struct file *exe_file;
1548
2b0143b5 1549 task = get_proc_task(d_inode(dentry));
925d1c40
MH		 1550 if (!task)
1551 return -ENOENT;
1552 mm = get_task_mm(task);
1553 put_task_struct(task);
1554 if (!mm)
1555 return -ENOENT;
1556 exe_file = get_mm_exe_file(mm);
1557 mmput(mm);
1558 if (exe_file) {
1559 *exe_path = exe_file->f_path;
1560 path_get(&exe_file->f_path);
1561 fput(exe_file);
1562 return 0;
1563 } else
1564 return -ENOENT;
1565}
1566
6e77137b 1567static const char *proc_pid_follow_link(struct dentry *dentry, void **cookie)
1da177e4 1568{
2b0143b5 1569 struct inode *inode = d_inode(dentry);
408ef013 1570 struct path path;
1da177e4
LT		 1571 int error = -EACCES;
1572
778c1144
EB		 1573 /* Are we allowed to snoop on the tasks file descriptors? */
1574 if (!proc_fd_access_allowed(inode))
1da177e4 1575 goto out;
1da177e4 1576
408ef013
CH		 1577 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1578 if (error)
1579 goto out;
1580
6e77137b 1581 nd_jump_link(&path);
408ef013 1582 return NULL;
1da177e4 1583out:
008b150a 1584 return ERR_PTR(error);
1da177e4
LT		 1585}
1586
3dcd25f3 1587static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1588{
e12ba74d 1589 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1590 char *pathname;
1da177e4
LT		 1591 int len;
1592
1593 if (!tmp)
1594 return -ENOMEM;
0c28f287 1595
7b2a69ba 1596 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1597 len = PTR_ERR(pathname);
1598 if (IS_ERR(pathname))
1da177e4 1599 goto out;
3dcd25f3 1600 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1601
1602 if (len > buflen)
1603 len = buflen;
3dcd25f3 1604 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1605 len = -EFAULT;
1606 out:
1607 free_page((unsigned long)tmp);
1608 return len;
1609}
1610
1611static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1612{
1613 int error = -EACCES;
2b0143b5 1614 struct inode *inode = d_inode(dentry);
3dcd25f3 1615 struct path path;
1da177e4 1616
778c1144
EB		 1617 /* Are we allowed to snoop on the tasks file descriptors? */
1618 if (!proc_fd_access_allowed(inode))
1da177e4 1619 goto out;
1da177e4 1620
7773fbc5 1621 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1622 if (error)
1623 goto out;
1624
3dcd25f3
JB		 1625 error = do_proc_readlink(&path, buffer, buflen);
1626 path_put(&path);
1da177e4 1627out:
1da177e4
LT		 1628 return error;
1629}
1630
faf60af1 1631const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1632 .readlink = proc_pid_readlink,
6d76fa58
LT		 1633 .follow_link = proc_pid_follow_link,
1634 .setattr = proc_setattr,
1da177e4
LT		 1635};
1636
28a6d671
EB		 1637
1638/* building an inode */
1639
6b4e306a 1640struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1641{
1642 struct inode * inode;
1643 struct proc_inode *ei;
c69e8d9c 1644 const struct cred *cred;
1da177e4 1645
28a6d671 1646 /* We need a new inode */
1da177e4 1647
28a6d671
EB		 1648 inode = new_inode(sb);
1649 if (!inode)
1650 goto out;
1651
1652 /* Common stuff */
1653 ei = PROC_I(inode);
85fe4025 1654 inode->i_ino = get_next_ino();
28a6d671 1655 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1656 inode->i_op = &proc_def_inode_operations;
1657
1658 /*
1659 * grab the reference to task.
1660 */
1a657f78 1661 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1662 if (!ei->pid)
1663 goto out_unlock;
1664
28a6d671 1665 if (task_dumpable(task)) {
c69e8d9c
DH		 1666 rcu_read_lock();
1667 cred = __task_cred(task);
1668 inode->i_uid = cred->euid;
1669 inode->i_gid = cred->egid;
1670 rcu_read_unlock();
1da177e4 1671 }
28a6d671
EB		 1672 security_task_to_inode(task, inode);
1673
1da177e4 1674out:
28a6d671
EB		 1675 return inode;
1676
1677out_unlock:
1678 iput(inode);
1679 return NULL;
1da177e4
LT		 1680}
1681
6b4e306a 1682int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1683{
2b0143b5 1684 struct inode *inode = d_inode(dentry);
28a6d671 1685 struct task_struct *task;
c69e8d9c 1686 const struct cred *cred;
0499680a 1687 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1688
28a6d671 1689 generic_fillattr(inode, stat);
1da177e4 1690
28a6d671 1691 rcu_read_lock();
dcb0f222
EB		 1692 stat->uid = GLOBAL_ROOT_UID;
1693 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1694 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1695 if (task) {
0499680a
VK		 1696 if (!has_pid_permissions(pid, task, 2)) {
1697 rcu_read_unlock();
1698 /*
1699 * This doesn't prevent learning whether PID exists,
1700 * it only makes getattr() consistent with readdir().
1701 */
1702 return -ENOENT;
1703 }
28a6d671
EB		 1704 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1705 task_dumpable(task)) {
c69e8d9c
DH		 1706 cred = __task_cred(task);
1707 stat->uid = cred->euid;
1708 stat->gid = cred->egid;
1da177e4
LT		 1709 }
1710 }
28a6d671 1711 rcu_read_unlock();
d6e71144 1712 return 0;
1da177e4
LT		 1713}
1714
1da177e4
LT		 1715/* dentry stuff */
1716
1717/*
1718 * Exceptional case: normally we are not allowed to unhash a busy
1719 * directory. In this case, however, we can do it - no aliasing problems
1720 * due to the way we treat inodes.
1721 *
1722 * Rewrite the inode's ownerships here because the owning task may have
1723 * performed a setuid(), etc.
99f89551
EB		 1724 *
1725 * Before the /proc/pid/status file was created the only way to read
1726 * the effective uid of a /process was to stat /proc/pid. Reading
1727 * /proc/pid/status is slow enough that procps and other packages
1728 * kept stating /proc/pid. To keep the rules in /proc simple I have
1729 * made this apply to all per process world readable and executable
1730 * directories.
1da177e4 1731 */
0b728e19 1732int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1733{
34286d66
NP		 1734 struct inode *inode;
1735 struct task_struct *task;
c69e8d9c
DH		 1736 const struct cred *cred;
1737
0b728e19 1738 if (flags & LOOKUP_RCU)
34286d66
NP		 1739 return -ECHILD;
1740
2b0143b5 1741 inode = d_inode(dentry);
34286d66
NP		 1742 task = get_proc_task(inode);
1743
99f89551
EB		 1744 if (task) {
1745 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1746 task_dumpable(task)) {
c69e8d9c
DH		 1747 rcu_read_lock();
1748 cred = __task_cred(task);
1749 inode->i_uid = cred->euid;
1750 inode->i_gid = cred->egid;
1751 rcu_read_unlock();
1da177e4 1752 } else {
dcb0f222
EB		 1753 inode->i_uid = GLOBAL_ROOT_UID;
1754 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1755 }
9ee8ab9f 1756 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1757 security_task_to_inode(task, inode);
99f89551 1758 put_task_struct(task);
1da177e4
LT		 1759 return 1;
1760 }
1da177e4
LT		 1761 return 0;
1762}
1763
d855a4b7
ON		 1764static inline bool proc_inode_is_dead(struct inode *inode)
1765{
1766 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1767}
1768
1dd704b6
DH		 1769int pid_delete_dentry(const struct dentry *dentry)
1770{
1771 /* Is the task we represent dead?
1772 * If so, then don't put the dentry on the lru list,
1773 * kill it immediately.
1774 */
2b0143b5 1775 return proc_inode_is_dead(d_inode(dentry));
1dd704b6
DH		 1776}
1777
6b4e306a 1778const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1779{
1780 .d_revalidate = pid_revalidate,
1781 .d_delete = pid_delete_dentry,
1782};
1783
1784/* Lookups */
1785
1c0d04c9
EB		 1786/*
1787 * Fill a directory entry.
1788 *
1789 * If possible create the dcache entry and derive our inode number and
1790 * file type from dcache entry.
1791 *
1792 * Since all of the proc inode numbers are dynamically generated, the inode
1793 * numbers do not exist until the inode is cache. This means creating the
1794 * the dcache entry in readdir is necessary to keep the inode numbers
1795 * reported by readdir in sync with the inode numbers reported
1796 * by stat.
1797 */
f0c3b509 1798bool proc_fill_cache(struct file *file, struct dir_context *ctx,
6b4e306a 1799 const char *name, int len,
c5141e6d 1800 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1801{
f0c3b509 1802 struct dentry *child, *dir = file->f_path.dentry;
1df98b8b 1803 struct qstr qname = QSTR_INIT(name, len);
61a28784 1804 struct inode *inode;
1df98b8b
AV		 1805 unsigned type;
1806 ino_t ino;
61a28784 1807
1df98b8b 1808 child = d_hash_and_lookup(dir, &qname);
61a28784 1809 if (!child) {
1df98b8b
AV		 1810 child = d_alloc(dir, &qname);
1811 if (!child)
1812 goto end_instantiate;
2b0143b5 1813 if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1df98b8b
AV		 1814 dput(child);
1815 goto end_instantiate;
61a28784
EB		 1816 }
1817 }
2b0143b5 1818 inode = d_inode(child);
147ce699
AV		 1819 ino = inode->i_ino;
1820 type = inode->i_mode >> 12;
61a28784 1821 dput(child);
f0c3b509 1822 return dir_emit(ctx, name, len, ino, type);
1df98b8b
AV		 1823
1824end_instantiate:
1825 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
61a28784
EB		 1826}
1827
640708a2
PE		 1828/*
1829 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1830 * which represent vma start and end addresses.
1831 */
1832static int dname_to_vma_addr(struct dentry *dentry,
1833 unsigned long *start, unsigned long *end)
1834{
1835 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1836 return -EINVAL;
1837
1838 return 0;
1839}
1840
0b728e19 1841static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1842{
1843 unsigned long vm_start, vm_end;
1844 bool exact_vma_exists = false;
1845 struct mm_struct *mm = NULL;
1846 struct task_struct *task;
1847 const struct cred *cred;
1848 struct inode *inode;
1849 int status = 0;
1850
0b728e19 1851 if (flags & LOOKUP_RCU)
640708a2
PE		 1852 return -ECHILD;
1853
2b0143b5 1854 inode = d_inode(dentry);
640708a2
PE		 1855 task = get_proc_task(inode);
1856 if (!task)
1857 goto out_notask;
1858
2344bec7
CW		 1859 mm = mm_access(task, PTRACE_MODE_READ);
1860 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1861 goto out;
1862
1863 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1864 down_read(&mm->mmap_sem);
1865 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1866 up_read(&mm->mmap_sem);
1867 }
1868
1869 mmput(mm);
1870
1871 if (exact_vma_exists) {
1872 if (task_dumpable(task)) {
1873 rcu_read_lock();
1874 cred = __task_cred(task);
1875 inode->i_uid = cred->euid;
1876 inode->i_gid = cred->egid;
1877 rcu_read_unlock();
1878 } else {
dcb0f222
EB		 1879 inode->i_uid = GLOBAL_ROOT_UID;
1880 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 1881 }
1882 security_task_to_inode(task, inode);
1883 status = 1;
1884 }
1885
1886out:
1887 put_task_struct(task);
1888
1889out_notask:
640708a2
PE		 1890 return status;
1891}
1892
1893static const struct dentry_operations tid_map_files_dentry_operations = {
1894 .d_revalidate = map_files_d_revalidate,
1895 .d_delete = pid_delete_dentry,
1896};
1897
1898static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1899{
1900 unsigned long vm_start, vm_end;
1901 struct vm_area_struct *vma;
1902 struct task_struct *task;
1903 struct mm_struct *mm;
1904 int rc;
1905
1906 rc = -ENOENT;
2b0143b5 1907 task = get_proc_task(d_inode(dentry));
640708a2
PE		 1908 if (!task)
1909 goto out;
1910
1911 mm = get_task_mm(task);
1912 put_task_struct(task);
1913 if (!mm)
1914 goto out;
1915
1916 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1917 if (rc)
1918 goto out_mmput;
1919
70335abb 1920 rc = -ENOENT;
640708a2
PE		 1921 down_read(&mm->mmap_sem);
1922 vma = find_exact_vma(mm, vm_start, vm_end);
1923 if (vma && vma->vm_file) {
1924 *path = vma->vm_file->f_path;
1925 path_get(path);
1926 rc = 0;
1927 }
1928 up_read(&mm->mmap_sem);
1929
1930out_mmput:
1931 mmput(mm);
1932out:
1933 return rc;
1934}
1935
1936struct map_files_info {
7b540d06 1937 fmode_t mode;
640708a2
PE		 1938 unsigned long len;
1939 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1940};
1941
bdb4d100
CO		 1942/*
1943 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1944 * symlinks may be used to bypass permissions on ancestor directories in the
1945 * path to the file in question.
1946 */
1947static const char *
1948proc_map_files_follow_link(struct dentry *dentry, void **cookie)
1949{
1950 if (!capable(CAP_SYS_ADMIN))
1951 return ERR_PTR(-EPERM);
1952
1953 return proc_pid_follow_link(dentry, NULL);
1954}
1955
1956/*
1957 * Identical to proc_pid_link_inode_operations except for follow_link()
1958 */
1959static const struct inode_operations proc_map_files_link_inode_operations = {
1960 .readlink = proc_pid_readlink,
1961 .follow_link = proc_map_files_follow_link,
1962 .setattr = proc_setattr,
1963};
1964
c52a47ac 1965static int
640708a2
PE		 1966proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1967 struct task_struct *task, const void *ptr)
1968{
7b540d06 1969 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 1970 struct proc_inode *ei;
1971 struct inode *inode;
1972
640708a2
PE		 1973 inode = proc_pid_make_inode(dir->i_sb, task);
1974 if (!inode)
c52a47ac 1975 return -ENOENT;
640708a2
PE		 1976
1977 ei = PROC_I(inode);
1978 ei->op.proc_get_link = proc_map_files_get_link;
1979
bdb4d100 1980 inode->i_op = &proc_map_files_link_inode_operations;
640708a2
PE		 1981 inode->i_size = 64;
1982 inode->i_mode = S_IFLNK;
1983
7b540d06 1984 if (mode & FMODE_READ)
640708a2 1985 inode->i_mode |= S_IRUSR;
7b540d06 1986 if (mode & FMODE_WRITE)
640708a2
PE		 1987 inode->i_mode |= S_IWUSR;
1988
1989 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1990 d_add(dentry, inode);
1991
c52a47ac 1992 return 0;
640708a2
PE		 1993}
1994
1995static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 1996 struct dentry *dentry, unsigned int flags)
640708a2
PE		 1997{
1998 unsigned long vm_start, vm_end;
1999 struct vm_area_struct *vma;
2000 struct task_struct *task;
c52a47ac 2001 int result;
640708a2
PE		 2002 struct mm_struct *mm;
2003
c52a47ac 2004 result = -ENOENT;
640708a2
PE		 2005 task = get_proc_task(dir);
2006 if (!task)
2007 goto out;
2008
c52a47ac 2009 result = -EACCES;
eb94cd96 2010 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2011 goto out_put_task;
2012
c52a47ac 2013 result = -ENOENT;
640708a2 2014 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2015 goto out_put_task;
640708a2
PE		 2016
2017 mm = get_task_mm(task);
2018 if (!mm)
eb94cd96 2019 goto out_put_task;
640708a2
PE		 2020
2021 down_read(&mm->mmap_sem);
2022 vma = find_exact_vma(mm, vm_start, vm_end);
2023 if (!vma)
2024 goto out_no_vma;
2025
05f56484
SK		 2026 if (vma->vm_file)
2027 result = proc_map_files_instantiate(dir, dentry, task,
2028 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 2029
2030out_no_vma:
2031 up_read(&mm->mmap_sem);
2032 mmput(mm);
640708a2
PE		 2033out_put_task:
2034 put_task_struct(task);
2035out:
c52a47ac 2036 return ERR_PTR(result);
640708a2
PE		 2037}
2038
2039static const struct inode_operations proc_map_files_inode_operations = {
2040 .lookup = proc_map_files_lookup,
2041 .permission = proc_fd_permission,
2042 .setattr = proc_setattr,
2043};
2044
2045static int
f0c3b509 2046proc_map_files_readdir(struct file *file, struct dir_context *ctx)
640708a2 2047{
640708a2
PE		 2048 struct vm_area_struct *vma;
2049 struct task_struct *task;
2050 struct mm_struct *mm;
f0c3b509
AV		 2051 unsigned long nr_files, pos, i;
2052 struct flex_array *fa = NULL;
2053 struct map_files_info info;
2054 struct map_files_info *p;
640708a2
PE		 2055 int ret;
2056
640708a2 2057 ret = -ENOENT;
f0c3b509 2058 task = get_proc_task(file_inode(file));
640708a2
PE		 2059 if (!task)
2060 goto out;
2061
2062 ret = -EACCES;
eb94cd96 2063 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2064 goto out_put_task;
2065
2066 ret = 0;
f0c3b509
AV		 2067 if (!dir_emit_dots(file, ctx))
2068 goto out_put_task;
640708a2 2069
f0c3b509
AV		 2070 mm = get_task_mm(task);
2071 if (!mm)
2072 goto out_put_task;
2073 down_read(&mm->mmap_sem);
640708a2 2074
f0c3b509 2075 nr_files = 0;
640708a2 2076
f0c3b509
AV		 2077 /*
2078 * We need two passes here:
2079 *
2080 * 1) Collect vmas of mapped files with mmap_sem taken
2081 * 2) Release mmap_sem and instantiate entries
2082 *
2083 * otherwise we get lockdep complained, since filldir()
2084 * routine might require mmap_sem taken in might_fault().
2085 */
640708a2 2086
f0c3b509
AV		 2087 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2088 if (vma->vm_file && ++pos > ctx->pos)
2089 nr_files++;
2090 }
2091
2092 if (nr_files) {
2093 fa = flex_array_alloc(sizeof(info), nr_files,
2094 GFP_KERNEL);
2095 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2096 GFP_KERNEL)) {
2097 ret = -ENOMEM;
2098 if (fa)
2099 flex_array_free(fa);
2100 up_read(&mm->mmap_sem);
2101 mmput(mm);
2102 goto out_put_task;
640708a2 2103 }
f0c3b509
AV		 2104 for (i = 0, vma = mm->mmap, pos = 2; vma;
2105 vma = vma->vm_next) {
2106 if (!vma->vm_file)
2107 continue;
2108 if (++pos <= ctx->pos)
2109 continue;
2110
2111 info.mode = vma->vm_file->f_mode;
2112 info.len = snprintf(info.name,
2113 sizeof(info.name), "%lx-%lx",
2114 vma->vm_start, vma->vm_end);
2115 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2116 BUG();
640708a2 2117 }
640708a2 2118 }
f0c3b509
AV		 2119 up_read(&mm->mmap_sem);
2120
2121 for (i = 0; i < nr_files; i++) {
2122 p = flex_array_get(fa, i);
2123 if (!proc_fill_cache(file, ctx,
2124 p->name, p->len,
2125 proc_map_files_instantiate,
2126 task,
2127 (void *)(unsigned long)p->mode))
2128 break;
2129 ctx->pos++;
640708a2 2130 }
f0c3b509
AV		 2131 if (fa)
2132 flex_array_free(fa);
2133 mmput(mm);
640708a2 2134
640708a2
PE		 2135out_put_task:
2136 put_task_struct(task);
2137out:
2138 return ret;
2139}
2140
2141static const struct file_operations proc_map_files_operations = {
2142 .read = generic_read_dir,
f0c3b509 2143 .iterate = proc_map_files_readdir,
640708a2
PE		 2144 .llseek = default_llseek,
2145};
2146
48f6a7a5
PE		 2147struct timers_private {
2148 struct pid *pid;
2149 struct task_struct *task;
2150 struct sighand_struct *sighand;
57b8015e 2151 struct pid_namespace *ns;
48f6a7a5
PE		 2152 unsigned long flags;
2153};
2154
2155static void *timers_start(struct seq_file *m, loff_t *pos)
2156{
2157 struct timers_private *tp = m->private;
2158
2159 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2160 if (!tp->task)
2161 return ERR_PTR(-ESRCH);
2162
2163 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2164 if (!tp->sighand)
2165 return ERR_PTR(-ESRCH);
2166
2167 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2168}
2169
2170static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2171{
2172 struct timers_private *tp = m->private;
2173 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2174}
2175
2176static void timers_stop(struct seq_file *m, void *v)
2177{
2178 struct timers_private *tp = m->private;
2179
2180 if (tp->sighand) {
2181 unlock_task_sighand(tp->task, &tp->flags);
2182 tp->sighand = NULL;
2183 }
2184
2185 if (tp->task) {
2186 put_task_struct(tp->task);
2187 tp->task = NULL;
2188 }
2189}
2190
2191static int show_timer(struct seq_file *m, void *v)
2192{
2193 struct k_itimer *timer;
57b8015e
PE		 2194 struct timers_private *tp = m->private;
2195 int notify;
cedbccab 2196 static const char * const nstr[] = {
57b8015e
PE		 2197 [SIGEV_SIGNAL] = "signal",
2198 [SIGEV_NONE] = "none",
2199 [SIGEV_THREAD] = "thread",
2200 };
48f6a7a5
PE		 2201
2202 timer = list_entry((struct list_head *)v, struct k_itimer, list);
57b8015e
PE		 2203 notify = timer->it_sigev_notify;
2204
48f6a7a5 2205 seq_printf(m, "ID: %d\n", timer->it_id);
25ce3191
JP		 2206 seq_printf(m, "signal: %d/%p\n",
2207 timer->sigq->info.si_signo,
2208 timer->sigq->info.si_value.sival_ptr);
57b8015e 2209 seq_printf(m, "notify: %s/%s.%d\n",
25ce3191
JP		 2210 nstr[notify & ~SIGEV_THREAD_ID],
2211 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2212 pid_nr_ns(timer->it_pid, tp->ns));
15ef0298 2213 seq_printf(m, "ClockID: %d\n", timer->it_clock);
48f6a7a5
PE		 2214
2215 return 0;
2216}
2217
2218static const struct seq_operations proc_timers_seq_ops = {
2219 .start = timers_start,
2220 .next = timers_next,
2221 .stop = timers_stop,
2222 .show = show_timer,
2223};
2224
2225static int proc_timers_open(struct inode *inode, struct file *file)
2226{
2227 struct timers_private *tp;
2228
2229 tp = __seq_open_private(file, &proc_timers_seq_ops,
2230 sizeof(struct timers_private));
2231 if (!tp)
2232 return -ENOMEM;
2233
2234 tp->pid = proc_pid(inode);
57b8015e 2235 tp->ns = inode->i_sb->s_fs_info;
48f6a7a5
PE		 2236 return 0;
2237}
2238
2239static const struct file_operations proc_timers_operations = {
2240 .open = proc_timers_open,
2241 .read = seq_read,
2242 .llseek = seq_lseek,
2243 .release = seq_release_private,
2244};
640708a2 2245
c52a47ac 2246static int proc_pident_instantiate(struct inode *dir,
c5141e6d 2247 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2248{
c5141e6d 2249 const struct pid_entry *p = ptr;
444ceed8
EB		 2250 struct inode *inode;
2251 struct proc_inode *ei;
444ceed8 2252
61a28784 2253 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2254 if (!inode)
2255 goto out;
2256
2257 ei = PROC_I(inode);
2258 inode->i_mode = p->mode;
2259 if (S_ISDIR(inode->i_mode))
bfe86848 2260 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2261 if (p->iop)
2262 inode->i_op = p->iop;
2263 if (p->fop)
2264 inode->i_fop = p->fop;
2265 ei->op = p->op;
fb045adb 2266 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2267 d_add(dentry, inode);
2268 /* Close the race of the process dying before we return the dentry */
0b728e19 2269 if (pid_revalidate(dentry, 0))
c52a47ac 2270 return 0;
444ceed8 2271out:
c52a47ac 2272 return -ENOENT;
444ceed8
EB		 2273}
2274
1da177e4
LT		 2275static struct dentry *proc_pident_lookup(struct inode *dir,
2276 struct dentry *dentry,
c5141e6d 2277 const struct pid_entry *ents,
7bcd6b0e 2278 unsigned int nents)
1da177e4 2279{
c52a47ac 2280 int error;
99f89551 2281 struct task_struct *task = get_proc_task(dir);
c5141e6d 2282 const struct pid_entry *p, *last;
1da177e4 2283
c52a47ac 2284 error = -ENOENT;
1da177e4 2285
99f89551
EB		 2286 if (!task)
2287 goto out_no_task;
1da177e4 2288
20cdc894
EB		 2289 /*
2290 * Yes, it does not scale. And it should not. Don't add
2291 * new entries into /proc/<tgid>/ without very good reasons.
2292 */
7bcd6b0e
EB		 2293 last = &ents[nents - 1];
2294 for (p = ents; p <= last; p++) {
1da177e4
LT		 2295 if (p->len != dentry->d_name.len)
2296 continue;
2297 if (!memcmp(dentry->d_name.name, p->name, p->len))
2298 break;
2299 }
7bcd6b0e 2300 if (p > last)
1da177e4
LT		 2301 goto out;
2302
444ceed8 2303 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2304out:
99f89551
EB		 2305 put_task_struct(task);
2306out_no_task:
c52a47ac 2307 return ERR_PTR(error);
1da177e4
LT		 2308}
2309
f0c3b509 2310static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
c5141e6d 2311 const struct pid_entry *ents, unsigned int nents)
28a6d671 2312{
f0c3b509
AV		 2313 struct task_struct *task = get_proc_task(file_inode(file));
2314 const struct pid_entry *p;
28a6d671 2315
28a6d671 2316 if (!task)
f0c3b509 2317 return -ENOENT;
28a6d671 2318
f0c3b509
AV		 2319 if (!dir_emit_dots(file, ctx))
2320 goto out;
2321
2322 if (ctx->pos >= nents + 2)
2323 goto out;
28a6d671 2324
f0c3b509
AV		 2325 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2326 if (!proc_fill_cache(file, ctx, p->name, p->len,
2327 proc_pident_instantiate, task, p))
2328 break;
2329 ctx->pos++;
2330 }
28a6d671 2331out:
61a28784 2332 put_task_struct(task);
f0c3b509 2333 return 0;
1da177e4
LT		 2334}
2335
28a6d671
EB		 2336#ifdef CONFIG_SECURITY
2337static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2338 size_t count, loff_t *ppos)
2339{
496ad9aa 2340 struct inode * inode = file_inode(file);
04ff9708 2341 char *p = NULL;
28a6d671
EB		 2342 ssize_t length;
2343 struct task_struct *task = get_proc_task(inode);
2344
28a6d671 2345 if (!task)
04ff9708 2346 return -ESRCH;
28a6d671
EB		 2347
2348 length = security_getprocattr(task,
2fddfeef 2349 (char*)file->f_path.dentry->d_name.name,
04ff9708 2350 &p);
28a6d671 2351 put_task_struct(task);
04ff9708
AV		 2352 if (length > 0)
2353 length = simple_read_from_buffer(buf, count, ppos, p, length);
2354 kfree(p);
28a6d671 2355 return length;
1da177e4
LT		 2356}
2357
28a6d671
EB		 2358static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2359 size_t count, loff_t *ppos)
2360{
496ad9aa 2361 struct inode * inode = file_inode(file);
28a6d671
EB		 2362 char *page;
2363 ssize_t length;
2364 struct task_struct *task = get_proc_task(inode);
2365
2366 length = -ESRCH;
2367 if (!task)
2368 goto out_no_task;
2369 if (count > PAGE_SIZE)
2370 count = PAGE_SIZE;
2371
2372 /* No partial writes. */
2373 length = -EINVAL;
2374 if (*ppos != 0)
2375 goto out;
2376
2377 length = -ENOMEM;
e12ba74d 2378 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2379 if (!page)
2380 goto out;
2381
2382 length = -EFAULT;
2383 if (copy_from_user(page, buf, count))
2384 goto out_free;
2385
107db7c7 2386 /* Guard against adverse ptrace interaction */
9b1bf12d 2387 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2388 if (length < 0)
2389 goto out_free;
2390
28a6d671 2391 length = security_setprocattr(task,
2fddfeef 2392 (char*)file->f_path.dentry->d_name.name,
28a6d671 2393 (void*)page, count);
9b1bf12d 2394 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2395out_free:
2396 free_page((unsigned long) page);
2397out:
2398 put_task_struct(task);
2399out_no_task:
2400 return length;
2401}
2402
00977a59 2403static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2404 .read = proc_pid_attr_read,
2405 .write = proc_pid_attr_write,
87df8424 2406 .llseek = generic_file_llseek,
28a6d671
EB		 2407};
2408
c5141e6d 2409static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2410 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2411 REG("prev", S_IRUGO, proc_pid_attr_operations),
2412 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2413 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2414 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2415 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2416};
2417
f0c3b509 2418static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
28a6d671 2419{
f0c3b509
AV		 2420 return proc_pident_readdir(file, ctx,
2421 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2422}
2423
00977a59 2424static const struct file_operations proc_attr_dir_operations = {
1da177e4 2425 .read = generic_read_dir,
f0c3b509 2426 .iterate = proc_attr_dir_readdir,
6038f373 2427 .llseek = default_llseek,
1da177e4
LT		 2428};
2429
72d9dcfc 2430static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2431 struct dentry *dentry, unsigned int flags)
28a6d671 2432{
7bcd6b0e
EB		 2433 return proc_pident_lookup(dir, dentry,
2434 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2435}
2436
c5ef1c42 2437static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2438 .lookup = proc_attr_dir_lookup,
99f89551 2439 .getattr = pid_getattr,
6d76fa58 2440 .setattr = proc_setattr,
1da177e4
LT		 2441};
2442
28a6d671
EB		 2443#endif
2444
698ba7b5 2445#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2446static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2447 size_t count, loff_t *ppos)
2448{
496ad9aa 2449 struct task_struct *task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2450 struct mm_struct *mm;
2451 char buffer[PROC_NUMBUF];
2452 size_t len;
2453 int ret;
2454
2455 if (!task)
2456 return -ESRCH;
2457
2458 ret = 0;
2459 mm = get_task_mm(task);
2460 if (mm) {
2461 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2462 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2463 MMF_DUMP_FILTER_SHIFT));
2464 mmput(mm);
2465 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2466 }
2467
2468 put_task_struct(task);
2469
2470 return ret;
2471}
2472
2473static ssize_t proc_coredump_filter_write(struct file *file,
2474 const char __user *buf,
2475 size_t count,
2476 loff_t *ppos)
2477{
2478 struct task_struct *task;
2479 struct mm_struct *mm;
3cb4a0bb
KH		 2480 unsigned int val;
2481 int ret;
2482 int i;
2483 unsigned long mask;
2484
774636e1
AD		 2485 ret = kstrtouint_from_user(buf, count, 0, &val);
2486 if (ret < 0)
2487 return ret;
3cb4a0bb
KH		 2488
2489 ret = -ESRCH;
496ad9aa 2490 task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2491 if (!task)
2492 goto out_no_task;
2493
3cb4a0bb
KH		 2494 mm = get_task_mm(task);
2495 if (!mm)
2496 goto out_no_mm;
2497
2498 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2499 if (val & mask)
2500 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2501 else
2502 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2503 }
2504
2505 mmput(mm);
2506 out_no_mm:
2507 put_task_struct(task);
2508 out_no_task:
774636e1
AD		 2509 if (ret < 0)
2510 return ret;
2511 return count;
3cb4a0bb
KH		 2512}
2513
2514static const struct file_operations proc_coredump_filter_operations = {
2515 .read = proc_coredump_filter_read,
2516 .write = proc_coredump_filter_write,
87df8424 2517 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2518};
2519#endif
2520
aba76fdb 2521#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2522static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
297c5d92 2523{
940389b8 2524 struct task_io_accounting acct = task->ioac;
5995477a 2525 unsigned long flags;
293eb1e7 2526 int result;
5995477a 2527
293eb1e7
VK		 2528 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2529 if (result)
2530 return result;
2531
2532 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2533 result = -EACCES;
2534 goto out_unlock;
2535 }
1d1221f3 2536
5995477a
AR		 2537 if (whole && lock_task_sighand(task, &flags)) {
2538 struct task_struct *t = task;
2539
2540 task_io_accounting_add(&acct, &task->signal->ioac);
2541 while_each_thread(task, t)
2542 task_io_accounting_add(&acct, &t->ioac);
2543
2544 unlock_task_sighand(task, &flags);
297c5d92 2545 }
25ce3191
JP		 2546 seq_printf(m,
2547 "rchar: %llu\n"
2548 "wchar: %llu\n"
2549 "syscr: %llu\n"
2550 "syscw: %llu\n"
2551 "read_bytes: %llu\n"
2552 "write_bytes: %llu\n"
2553 "cancelled_write_bytes: %llu\n",
2554 (unsigned long long)acct.rchar,
2555 (unsigned long long)acct.wchar,
2556 (unsigned long long)acct.syscr,
2557 (unsigned long long)acct.syscw,
2558 (unsigned long long)acct.read_bytes,
2559 (unsigned long long)acct.write_bytes,
2560 (unsigned long long)acct.cancelled_write_bytes);
2561 result = 0;
2562
293eb1e7
VK		 2563out_unlock:
2564 mutex_unlock(&task->signal->cred_guard_mutex);
2565 return result;
297c5d92
AR		 2566}
2567
19aadc98
AD		 2568static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2569 struct pid *pid, struct task_struct *task)
297c5d92 2570{
19aadc98 2571 return do_io_accounting(task, m, 0);
aba76fdb 2572}
297c5d92 2573
19aadc98
AD		 2574static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2575 struct pid *pid, struct task_struct *task)
297c5d92 2576{
19aadc98 2577 return do_io_accounting(task, m, 1);
297c5d92
AR		 2578}
2579#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2580
22d917d8
EB		 2581#ifdef CONFIG_USER_NS
2582static int proc_id_map_open(struct inode *inode, struct file *file,
ccf94f1b 2583 const struct seq_operations *seq_ops)
22d917d8
EB		 2584{
2585 struct user_namespace *ns = NULL;
2586 struct task_struct *task;
2587 struct seq_file *seq;
2588 int ret = -EINVAL;
2589
2590 task = get_proc_task(inode);
2591 if (task) {
2592 rcu_read_lock();
2593 ns = get_user_ns(task_cred_xxx(task, user_ns));
2594 rcu_read_unlock();
2595 put_task_struct(task);
2596 }
2597 if (!ns)
2598 goto err;
2599
2600 ret = seq_open(file, seq_ops);
2601 if (ret)
2602 goto err_put_ns;
2603
2604 seq = file->private_data;
2605 seq->private = ns;
2606
2607 return 0;
2608err_put_ns:
2609 put_user_ns(ns);
2610err:
2611 return ret;
2612}
2613
2614static int proc_id_map_release(struct inode *inode, struct file *file)
2615{
2616 struct seq_file *seq = file->private_data;
2617 struct user_namespace *ns = seq->private;
2618 put_user_ns(ns);
2619 return seq_release(inode, file);
2620}
2621
2622static int proc_uid_map_open(struct inode *inode, struct file *file)
2623{
2624 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2625}
2626
2627static int proc_gid_map_open(struct inode *inode, struct file *file)
2628{
2629 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2630}
2631
f76d207a
EB		 2632static int proc_projid_map_open(struct inode *inode, struct file *file)
2633{
2634 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2635}
2636
22d917d8
EB		 2637static const struct file_operations proc_uid_map_operations = {
2638 .open = proc_uid_map_open,
2639 .write = proc_uid_map_write,
2640 .read = seq_read,
2641 .llseek = seq_lseek,
2642 .release = proc_id_map_release,
2643};
2644
2645static const struct file_operations proc_gid_map_operations = {
2646 .open = proc_gid_map_open,
2647 .write = proc_gid_map_write,
2648 .read = seq_read,
2649 .llseek = seq_lseek,
2650 .release = proc_id_map_release,
2651};
f76d207a
EB		 2652
2653static const struct file_operations proc_projid_map_operations = {
2654 .open = proc_projid_map_open,
2655 .write = proc_projid_map_write,
2656 .read = seq_read,
2657 .llseek = seq_lseek,
2658 .release = proc_id_map_release,
2659};
9cc46516
EB		 2660
2661static int proc_setgroups_open(struct inode *inode, struct file *file)
2662{
2663 struct user_namespace *ns = NULL;
2664 struct task_struct *task;
2665 int ret;
2666
2667 ret = -ESRCH;
2668 task = get_proc_task(inode);
2669 if (task) {
2670 rcu_read_lock();
2671 ns = get_user_ns(task_cred_xxx(task, user_ns));
2672 rcu_read_unlock();
2673 put_task_struct(task);
2674 }
2675 if (!ns)
2676 goto err;
2677
2678 if (file->f_mode & FMODE_WRITE) {
2679 ret = -EACCES;
2680 if (!ns_capable(ns, CAP_SYS_ADMIN))
2681 goto err_put_ns;
2682 }
2683
2684 ret = single_open(file, &proc_setgroups_show, ns);
2685 if (ret)
2686 goto err_put_ns;
2687
2688 return 0;
2689err_put_ns:
2690 put_user_ns(ns);
2691err:
2692 return ret;
2693}
2694
2695static int proc_setgroups_release(struct inode *inode, struct file *file)
2696{
2697 struct seq_file *seq = file->private_data;
2698 struct user_namespace *ns = seq->private;
2699 int ret = single_release(inode, file);
2700 put_user_ns(ns);
2701 return ret;
2702}
2703
2704static const struct file_operations proc_setgroups_operations = {
2705 .open = proc_setgroups_open,
2706 .write = proc_setgroups_write,
2707 .read = seq_read,
2708 .llseek = seq_lseek,
2709 .release = proc_setgroups_release,
2710};
22d917d8
EB		 2711#endif /* CONFIG_USER_NS */
2712
47830723
KC		 2713static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2714 struct pid *pid, struct task_struct *task)
2715{
a9712bc1
AV		 2716 int err = lock_trace(task);
2717 if (!err) {
2718 seq_printf(m, "%08x\n", task->personality);
2719 unlock_trace(task);
2720 }
2721 return err;
47830723
KC		 2722}
2723
28a6d671
EB		 2724/*
2725 * Thread groups
2726 */
00977a59 2727static const struct file_operations proc_task_operations;
c5ef1c42 2728static const struct inode_operations proc_task_inode_operations;
20cdc894 2729
c5141e6d 2730static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2731 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2732 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2 2733 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
631f9c18 2734 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2735 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2736#ifdef CONFIG_NET
631f9c18 2737 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2738#endif
631f9c18 2739 REG("environ", S_IRUSR, proc_environ_operations),
f9ea536e 2740 ONE("auxv", S_IRUSR, proc_pid_auxv),
631f9c18 2741 ONE("status", S_IRUGO, proc_pid_status),
35a35046 2742 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 2743 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2744#ifdef CONFIG_SCHED_DEBUG
631f9c18 2745 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2746#endif
2747#ifdef CONFIG_SCHED_AUTOGROUP
2748 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2749#endif
4614a696 2750 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2751#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 2752 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2753#endif
c2c0bb44 2754 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 2755 ONE("stat", S_IRUGO, proc_tgid_stat),
2756 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2757 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2758#ifdef CONFIG_NUMA
b7643757 2759 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2760#endif
631f9c18
AD		 2761 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2762 LNK("cwd", proc_cwd_link),
2763 LNK("root", proc_root_link),
2764 LNK("exe", proc_exe_link),
2765 REG("mounts", S_IRUGO, proc_mounts_operations),
2766 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2767 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2768#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2769 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2770 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
32ed74a4 2771 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2772#endif
2773#ifdef CONFIG_SECURITY
631f9c18 2774 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2775#endif
2776#ifdef CONFIG_KALLSYMS
edfcd606 2777 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2778#endif
2ec220e2 2779#ifdef CONFIG_STACKTRACE
35a35046 2780 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 2781#endif
5968cece 2782#ifdef CONFIG_SCHED_INFO
f6e826ca 2783 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2784#endif
9745512c 2785#ifdef CONFIG_LATENCYTOP
631f9c18 2786 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2787#endif
8793d854 2788#ifdef CONFIG_PROC_PID_CPUSET
52de4779 2789 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 2790#endif
2791#ifdef CONFIG_CGROUPS
006f4ac4 2792 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 2793#endif
6ba51e37 2794 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 2795 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 2796 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2797#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2798 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2799 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2800#endif
f4f154fd 2801#ifdef CONFIG_FAULT_INJECTION
631f9c18 2802 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2803#endif
698ba7b5 2804#ifdef CONFIG_ELF_CORE
631f9c18 2805 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2806#endif
aba76fdb 2807#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2808 ONE("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2809#endif
f133ecca 2810#ifdef CONFIG_HARDWALL
d962c144 2811 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 2812#endif
22d917d8
EB		 2813#ifdef CONFIG_USER_NS
2814 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2815 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 2816 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 2817 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 2818#endif
48f6a7a5
PE		 2819#ifdef CONFIG_CHECKPOINT_RESTORE
2820 REG("timers", S_IRUGO, proc_timers_operations),
2821#endif
28a6d671 2822};
1da177e4 2823
f0c3b509 2824static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
1da177e4 2825{
f0c3b509
AV		 2826 return proc_pident_readdir(file, ctx,
2827 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2828}
2829
00977a59 2830static const struct file_operations proc_tgid_base_operations = {
1da177e4 2831 .read = generic_read_dir,
f0c3b509 2832 .iterate = proc_tgid_base_readdir,
6038f373 2833 .llseek = default_llseek,
1da177e4
LT		 2834};
2835
00cd8dd3
AV		 2836static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2837{
7bcd6b0e
EB		 2838 return proc_pident_lookup(dir, dentry,
2839 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2840}
2841
c5ef1c42 2842static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2843 .lookup = proc_tgid_base_lookup,
99f89551 2844 .getattr = pid_getattr,
6d76fa58 2845 .setattr = proc_setattr,
0499680a 2846 .permission = proc_pid_permission,
1da177e4 2847};
1da177e4 2848
60347f67 2849static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2850{
48e6484d 2851 struct dentry *dentry, *leader, *dir;
8578cea7 2852 char buf[PROC_NUMBUF];
48e6484d
EB		 2853 struct qstr name;
2854
2855 name.name = buf;
60347f67 2856 name.len = snprintf(buf, sizeof(buf), "%d", pid);
4f522a24 2857 /* no ->d_hash() rejects on procfs */
60347f67 2858 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2859 if (dentry) {
bbd51924 2860 d_invalidate(dentry);
48e6484d
EB		 2861 dput(dentry);
2862 }
1da177e4 2863
c35a7f18
ON		 2864 if (pid == tgid)
2865 return;
2866
48e6484d 2867 name.name = buf;
60347f67
PE		 2868 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2869 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2870 if (!leader)
2871 goto out;
1da177e4 2872
48e6484d
EB		 2873 name.name = "task";
2874 name.len = strlen(name.name);
2875 dir = d_hash_and_lookup(leader, &name);
2876 if (!dir)
2877 goto out_put_leader;
2878
2879 name.name = buf;
60347f67 2880 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2881 dentry = d_hash_and_lookup(dir, &name);
2882 if (dentry) {
bbd51924 2883 d_invalidate(dentry);
48e6484d 2884 dput(dentry);
1da177e4 2885 }
48e6484d
EB		 2886
2887 dput(dir);
2888out_put_leader:
2889 dput(leader);
2890out:
2891 return;
1da177e4
LT		 2892}
2893
0895e91d
RD		 2894/**
2895 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2896 * @task: task that should be flushed.
2897 *
2898 * When flushing dentries from proc, one needs to flush them from global
60347f67 2899 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2900 * in. This call is supposed to do all of this job.
2901 *
2902 * Looks in the dcache for
2903 * /proc/@pid
2904 * /proc/@tgid/task/@pid
2905 * if either directory is present flushes it and all of it'ts children
2906 * from the dcache.
2907 *
2908 * It is safe and reasonable to cache /proc entries for a task until
2909 * that task exits. After that they just clog up the dcache with
2910 * useless entries, possibly causing useful dcache entries to be
2911 * flushed instead. This routine is proved to flush those useless
2912 * dcache entries at process exit time.
2913 *
2914 * NOTE: This routine is just an optimization so it does not guarantee
2915 * that no dcache entries will exist at process exit time it
2916 * just makes it very unlikely that any will persist.
60347f67
PE		 2917 */
2918
2919void proc_flush_task(struct task_struct *task)
2920{
9fcc2d15 2921 int i;
9b4d1cbe 2922 struct pid *pid, *tgid;
130f77ec
PE		 2923 struct upid *upid;
2924
130f77ec 2925 pid = task_pid(task);
9b4d1cbe 2926 tgid = task_tgid(task);
130f77ec 2927
9fcc2d15 2928 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2929 upid = &pid->numbers[i];
2930 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2931 tgid->numbers[i].nr);
130f77ec 2932 }
60347f67
PE		 2933}
2934
c52a47ac
AV		 2935static int proc_pid_instantiate(struct inode *dir,
2936 struct dentry * dentry,
2937 struct task_struct *task, const void *ptr)
444ceed8 2938{
444ceed8
EB		 2939 struct inode *inode;
2940
61a28784 2941 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2942 if (!inode)
2943 goto out;
2944
2945 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2946 inode->i_op = &proc_tgid_base_inode_operations;
2947 inode->i_fop = &proc_tgid_base_operations;
2948 inode->i_flags|=S_IMMUTABLE;
aed54175 2949
bfe86848
MS		 2950 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2951 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 2952
fb045adb 2953 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2954
2955 d_add(dentry, inode);
2956 /* Close the race of the process dying before we return the dentry */
0b728e19 2957 if (pid_revalidate(dentry, 0))
c52a47ac 2958 return 0;
444ceed8 2959out:
c52a47ac 2960 return -ENOENT;
444ceed8
EB		 2961}
2962
00cd8dd3 2963struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 2964{
335eb531 2965 int result = -ENOENT;
1da177e4 2966 struct task_struct *task;
1da177e4 2967 unsigned tgid;
b488893a 2968 struct pid_namespace *ns;
1da177e4 2969
dbcdb504 2970 tgid = name_to_int(&dentry->d_name);
1da177e4
LT		 2971 if (tgid == ~0U)
2972 goto out;
2973
b488893a 2974 ns = dentry->d_sb->s_fs_info;
de758734 2975 rcu_read_lock();
b488893a 2976 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2977 if (task)
2978 get_task_struct(task);
de758734 2979 rcu_read_unlock();
1da177e4
LT		 2980 if (!task)
2981 goto out;
2982
444ceed8 2983 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 2984 put_task_struct(task);
1da177e4 2985out:
c52a47ac 2986 return ERR_PTR(result);
1da177e4
LT		 2987}
2988
1da177e4 2989/*
0804ef4b 2990 * Find the first task with tgid >= tgid
0bc58a91 2991 *
1da177e4 2992 */
19fd4bb2
EB		 2993struct tgid_iter {
2994 unsigned int tgid;
0804ef4b 2995 struct task_struct *task;
19fd4bb2
EB		 2996};
2997static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2998{
0804ef4b 2999 struct pid *pid;
1da177e4 3000
19fd4bb2
EB		 3001 if (iter.task)
3002 put_task_struct(iter.task);
454cc105 3003 rcu_read_lock();
0804ef4b 3004retry:
19fd4bb2
EB		 3005 iter.task = NULL;
3006 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3007 if (pid) {
19fd4bb2
EB		 3008 iter.tgid = pid_nr_ns(pid, ns);
3009 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3010 /* What we to know is if the pid we have find is the
3011 * pid of a thread_group_leader. Testing for task
3012 * being a thread_group_leader is the obvious thing
3013 * todo but there is a window when it fails, due to
3014 * the pid transfer logic in de_thread.
3015 *
3016 * So we perform the straight forward test of seeing
3017 * if the pid we have found is the pid of a thread
3018 * group leader, and don't worry if the task we have
3019 * found doesn't happen to be a thread group leader.
3020 * As we don't care in the case of readdir.
3021 */
19fd4bb2
EB		 3022 if (!iter.task || !has_group_leader_pid(iter.task)) {
3023 iter.tgid += 1;
0804ef4b 3024 goto retry;
19fd4bb2
EB		 3025 }
3026 get_task_struct(iter.task);
0bc58a91 3027 }
454cc105 3028 rcu_read_unlock();
19fd4bb2 3029 return iter;
1da177e4
LT		 3030}
3031
0097875b 3032#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
0804ef4b 3033
1da177e4 3034/* for the /proc/ directory itself, after non-process stuff has been done */
f0c3b509 3035int proc_pid_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3036{
19fd4bb2 3037 struct tgid_iter iter;
3aa3377f 3038 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
f0c3b509 3039 loff_t pos = ctx->pos;
1da177e4 3040
021ada7d 3041 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
f0c3b509 3042 return 0;
1da177e4 3043
0097875b 3044 if (pos == TGID_OFFSET - 2) {
2b0143b5 3045 struct inode *inode = d_inode(ns->proc_self);
db963164 3046 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
f0c3b509 3047 return 0;
0097875b
EB		 3048 ctx->pos = pos = pos + 1;
3049 }
3050 if (pos == TGID_OFFSET - 1) {
2b0143b5 3051 struct inode *inode = d_inode(ns->proc_thread_self);
0097875b
EB		 3052 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3053 return 0;
3054 ctx->pos = pos = pos + 1;
021ada7d 3055 }
0097875b 3056 iter.tgid = pos - TGID_OFFSET;
19fd4bb2 3057 iter.task = NULL;
19fd4bb2
EB		 3058 for (iter = next_tgid(ns, iter);
3059 iter.task;
3060 iter.tgid += 1, iter = next_tgid(ns, iter)) {
f0c3b509
AV		 3061 char name[PROC_NUMBUF];
3062 int len;
3063 if (!has_pid_permissions(ns, iter.task, 2))
3064 continue;
0499680a 3065
f0c3b509
AV		 3066 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3067 ctx->pos = iter.tgid + TGID_OFFSET;
3068 if (!proc_fill_cache(file, ctx, name, len,
3069 proc_pid_instantiate, iter.task, NULL)) {
19fd4bb2 3070 put_task_struct(iter.task);
f0c3b509 3071 return 0;
1da177e4 3072 }
0bc58a91 3073 }
f0c3b509 3074 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
0bc58a91
EB		 3075 return 0;
3076}
1da177e4 3077
28a6d671
EB		 3078/*
3079 * Tasks
3080 */
c5141e6d 3081static const struct pid_entry tid_base_stuff[] = {
631f9c18 3082 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3083 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3084 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
6ba8ed79
EB		 3085#ifdef CONFIG_NET
3086 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3087#endif
631f9c18 3088 REG("environ", S_IRUSR, proc_environ_operations),
f9ea536e 3089 ONE("auxv", S_IRUSR, proc_pid_auxv),
631f9c18 3090 ONE("status", S_IRUGO, proc_pid_status),
35a35046 3091 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 3092 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3093#ifdef CONFIG_SCHED_DEBUG
631f9c18 3094 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3095#endif
4614a696 3096 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3097#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 3098 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3099#endif
c2c0bb44 3100 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 3101 ONE("stat", S_IRUGO, proc_tid_stat),
3102 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3103 REG("maps", S_IRUGO, proc_tid_maps_operations),
2e13ba54 3104#ifdef CONFIG_PROC_CHILDREN
81841161
CG		 3105 REG("children", S_IRUGO, proc_tid_children_operations),
3106#endif
28a6d671 3107#ifdef CONFIG_NUMA
b7643757 3108 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3109#endif
631f9c18
AD		 3110 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3111 LNK("cwd", proc_cwd_link),
3112 LNK("root", proc_root_link),
3113 LNK("exe", proc_exe_link),
3114 REG("mounts", S_IRUGO, proc_mounts_operations),
3115 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3116#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3117 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3118 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
32ed74a4 3119 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3120#endif
3121#ifdef CONFIG_SECURITY
631f9c18 3122 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3123#endif
3124#ifdef CONFIG_KALLSYMS
edfcd606 3125 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3126#endif
2ec220e2 3127#ifdef CONFIG_STACKTRACE
35a35046 3128 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 3129#endif
5968cece 3130#ifdef CONFIG_SCHED_INFO
f6e826ca 3131 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3132#endif
9745512c 3133#ifdef CONFIG_LATENCYTOP
631f9c18 3134 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3135#endif
8793d854 3136#ifdef CONFIG_PROC_PID_CPUSET
52de4779 3137 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 3138#endif
3139#ifdef CONFIG_CGROUPS
006f4ac4 3140 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 3141#endif
6ba51e37 3142 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 3143 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3144 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3145#ifdef CONFIG_AUDITSYSCALL
631f9c18 3146 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3147 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3148#endif
f4f154fd 3149#ifdef CONFIG_FAULT_INJECTION
631f9c18 3150 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3151#endif
297c5d92 3152#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3153 ONE("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3154#endif
f133ecca 3155#ifdef CONFIG_HARDWALL
d962c144 3156 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 3157#endif
22d917d8
EB		 3158#ifdef CONFIG_USER_NS
3159 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3160 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3161 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3162 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3163#endif
28a6d671
EB		 3164};
3165
f0c3b509 3166static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
28a6d671 3167{
f0c3b509
AV		 3168 return proc_pident_readdir(file, ctx,
3169 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3170}
3171
00cd8dd3
AV		 3172static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3173{
7bcd6b0e
EB		 3174 return proc_pident_lookup(dir, dentry,
3175 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3176}
3177
00977a59 3178static const struct file_operations proc_tid_base_operations = {
28a6d671 3179 .read = generic_read_dir,
f0c3b509 3180 .iterate = proc_tid_base_readdir,
6038f373 3181 .llseek = default_llseek,
28a6d671
EB		 3182};
3183
c5ef1c42 3184static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3185 .lookup = proc_tid_base_lookup,
3186 .getattr = pid_getattr,
3187 .setattr = proc_setattr,
3188};
3189
c52a47ac 3190static int proc_task_instantiate(struct inode *dir,
c5141e6d 3191 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 3192{
444ceed8 3193 struct inode *inode;
61a28784 3194 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3195
3196 if (!inode)
3197 goto out;
3198 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3199 inode->i_op = &proc_tid_base_inode_operations;
3200 inode->i_fop = &proc_tid_base_operations;
3201 inode->i_flags|=S_IMMUTABLE;
aed54175 3202
bfe86848
MS		 3203 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3204 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3205
fb045adb 3206 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3207
3208 d_add(dentry, inode);
3209 /* Close the race of the process dying before we return the dentry */
0b728e19 3210 if (pid_revalidate(dentry, 0))
c52a47ac 3211 return 0;
444ceed8 3212out:
c52a47ac 3213 return -ENOENT;
444ceed8
EB		 3214}
3215
00cd8dd3 3216static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671 3217{
c52a47ac 3218 int result = -ENOENT;
28a6d671
EB		 3219 struct task_struct *task;
3220 struct task_struct *leader = get_proc_task(dir);
28a6d671 3221 unsigned tid;
b488893a 3222 struct pid_namespace *ns;
28a6d671
EB		 3223
3224 if (!leader)
3225 goto out_no_task;
3226
dbcdb504 3227 tid = name_to_int(&dentry->d_name);
28a6d671
EB		 3228 if (tid == ~0U)
3229 goto out;
3230
b488893a 3231 ns = dentry->d_sb->s_fs_info;
28a6d671 3232 rcu_read_lock();
b488893a 3233 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3234 if (task)
3235 get_task_struct(task);
3236 rcu_read_unlock();
3237 if (!task)
3238 goto out;
bac0abd6 3239 if (!same_thread_group(leader, task))
28a6d671
EB		 3240 goto out_drop_task;
3241
444ceed8 3242 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3243out_drop_task:
3244 put_task_struct(task);
3245out:
3246 put_task_struct(leader);
3247out_no_task:
c52a47ac 3248 return ERR_PTR(result);
28a6d671
EB		 3249}
3250
0bc58a91
EB		 3251/*
3252 * Find the first tid of a thread group to return to user space.
3253 *
3254 * Usually this is just the thread group leader, but if the users
3255 * buffer was too small or there was a seek into the middle of the
3256 * directory we have more work todo.
3257 *
3258 * In the case of a short read we start with find_task_by_pid.
3259 *
3260 * In the case of a seek we start with the leader and walk nr
3261 * threads past it.
3262 */
9f6e963f
ON		 3263static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3264 struct pid_namespace *ns)
0bc58a91 3265{
d855a4b7 3266 struct task_struct *pos, *task;
9f6e963f
ON		 3267 unsigned long nr = f_pos;
3268
3269 if (nr != f_pos) /* 32bit overflow? */
3270 return NULL;
1da177e4 3271
cc288738 3272 rcu_read_lock();
d855a4b7
ON		 3273 task = pid_task(pid, PIDTYPE_PID);
3274 if (!task)
3275 goto fail;
3276
3277 /* Attempt to start with the tid of a thread */
9f6e963f 3278 if (tid && nr) {
b488893a 3279 pos = find_task_by_pid_ns(tid, ns);
d855a4b7 3280 if (pos && same_thread_group(pos, task))
a872ff0c 3281 goto found;
0bc58a91 3282 }
1da177e4 3283
0bc58a91 3284 /* If nr exceeds the number of threads there is nothing todo */
9f6e963f 3285 if (nr >= get_nr_threads(task))
c986c14a 3286 goto fail;
1da177e4 3287
a872ff0c
ON		 3288 /* If we haven't found our starting place yet start
3289 * with the leader and walk nr threads forward.
0bc58a91 3290 */
d855a4b7 3291 pos = task = task->group_leader;
c986c14a 3292 do {
9f6e963f 3293 if (!nr--)
c986c14a 3294 goto found;
d855a4b7 3295 } while_each_thread(task, pos);
c986c14a
ON		 3296fail:
3297 pos = NULL;
3298 goto out;
a872ff0c
ON		 3299found:
3300 get_task_struct(pos);
3301out:
cc288738 3302 rcu_read_unlock();
0bc58a91
EB		 3303 return pos;
3304}
3305
3306/*
3307 * Find the next thread in the thread list.
3308 * Return NULL if there is an error or no next thread.
3309 *
3310 * The reference to the input task_struct is released.
3311 */
3312static struct task_struct *next_tid(struct task_struct *start)
3313{
c1df7fb8 3314 struct task_struct *pos = NULL;
cc288738 3315 rcu_read_lock();
c1df7fb8 3316 if (pid_alive(start)) {
0bc58a91 3317 pos = next_thread(start);
c1df7fb8
ON		 3318 if (thread_group_leader(pos))
3319 pos = NULL;
3320 else
3321 get_task_struct(pos);
3322 }
cc288738 3323 rcu_read_unlock();
0bc58a91
EB		 3324 put_task_struct(start);
3325 return pos;
1da177e4
LT		 3326}
3327
3328/* for the /proc/TGID/task/ directories */
f0c3b509 3329static int proc_task_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3330{
d855a4b7
ON		 3331 struct inode *inode = file_inode(file);
3332 struct task_struct *task;
b488893a 3333 struct pid_namespace *ns;
f0c3b509 3334 int tid;
1da177e4 3335
d855a4b7 3336 if (proc_inode_is_dead(inode))
f0c3b509 3337 return -ENOENT;
1da177e4 3338
f0c3b509 3339 if (!dir_emit_dots(file, ctx))
d855a4b7 3340 return 0;
1da177e4 3341
0bc58a91
EB		 3342 /* f_version caches the tgid value that the last readdir call couldn't
3343 * return. lseek aka telldir automagically resets f_version to 0.
3344 */
3aa3377f 3345 ns = inode->i_sb->s_fs_info;
f0c3b509
AV		 3346 tid = (int)file->f_version;
3347 file->f_version = 0;
d855a4b7 3348 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
0bc58a91 3349 task;
f0c3b509
AV		 3350 task = next_tid(task), ctx->pos++) {
3351 char name[PROC_NUMBUF];
3352 int len;
b488893a 3353 tid = task_pid_nr_ns(task, ns);
f0c3b509
AV		 3354 len = snprintf(name, sizeof(name), "%d", tid);
3355 if (!proc_fill_cache(file, ctx, name, len,
3356 proc_task_instantiate, task, NULL)) {
0bc58a91
EB		 3357 /* returning this tgid failed, save it as the first
3358 * pid for the next readir call */
f0c3b509 3359 file->f_version = (u64)tid;
0bc58a91 3360 put_task_struct(task);
1da177e4 3361 break;
0bc58a91 3362 }
1da177e4 3363 }
d855a4b7 3364
f0c3b509 3365 return 0;
1da177e4 3366}
6e66b52b
EB		 3367
3368static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3369{
2b0143b5 3370 struct inode *inode = d_inode(dentry);
99f89551 3371 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3372 generic_fillattr(inode, stat);
3373
99f89551 3374 if (p) {
99f89551 3375 stat->nlink += get_nr_threads(p);
99f89551 3376 put_task_struct(p);
6e66b52b
EB		 3377 }
3378
3379 return 0;
3380}
28a6d671 3381
c5ef1c42 3382static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3383 .lookup = proc_task_lookup,
3384 .getattr = proc_task_getattr,
3385 .setattr = proc_setattr,
0499680a 3386 .permission = proc_pid_permission,
28a6d671
EB		 3387};
3388
00977a59 3389static const struct file_operations proc_task_operations = {
28a6d671 3390 .read = generic_read_dir,
f0c3b509 3391 .iterate = proc_task_readdir,
6038f373 3392 .llseek = default_llseek,
28a6d671 3393};
Linux 6.x block layer and io_uring tree(s)